LiteLLM · Python benchmark measurements
Operated by the Busbar team; Busbar is also an entrant. Gateway overhead against a local mock is not end-to-end model latency. Methodology and limitations.
Python · recorded build: ghcr.io/berriai/litellm:v1.108.3. Measured: .
JSON · Markdown · CSV · Project source · Interactive dashboard
Each row below names its protocol pair. A representative dashboard cell may differ between gateways; compare an identical pair before interpreting a ranking. Unavailable metrics retain their reason; below-resolution readings are not exact zeroes.
Published measurements
| Ingress → upstream | Metric | Value | State / reason | p99 bound (ms) |
|---|---|---|---|---|
| Cohere → Cohere | Added latency p50 (µs) | 11,850 | Valid · n=503 gateway / 264,219 direct | — |
| Cohere → Cohere | Added latency p99 (µs) | 12,623 | Valid · n=503 gateway / 264,219 direct | — |
| Cohere → Cohere | Gateway p99 at concurrency 1 (µs) | 12,651 | Valid · n=503 | — |
| Cohere → Cohere | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=264,219 | — |
| Cohere → Cohere | CPU per request (µs) | 12,063.5 | Valid | — |
| Cohere → Cohere | Requests per CPU-second | 82.89 | Valid | — |
| Cohere → Cohere | Cost window concurrency | 8 | Valid | — |
| Cohere → Cohere | Cost window core utilisation | 0.25 | Valid | — |
| Cohere → Cohere | Cost window successful requests | 504 | Valid | — |
| Cohere → Cohere | Cost window throughput (req/s) | 83 | Valid | — |
| Cohere → Cohere | Cost window threads | 6 | Valid | — |
| Cohere → Cohere | Non-voluntary context switches per request | 0.052 | Valid | — |
| Cohere → Cohere | Cost window major page faults | 0 | Valid | — |
| Cohere → Cohere | Added time to first event p50 (µs) | — | Untestable | — |
| Cohere → Cohere | Added time to first event p99 (µs) | — | Untestable | — |
| Cohere → Cohere | Added inter-frame gap p50 (µs) | — | Untestable | — |
| Cohere → Cohere | Added inter-frame gap p99 (µs) | — | Untestable | — |
| Cohere → Cohere | Streams sustained (frames/s) | — | Untestable | — |
| Cohere → Cohere | Streams sustained | — | Untestable | — |
| Cohere → Cohere | Idle RSS (MiB) | 592.8 | Valid | — |
| Cohere → Cohere | Steady-state RSS (MiB) | 663.3 | Valid | — |
| Cohere → Cohere | Recovered RSS (MiB) | 663.3 | Valid | — |
| Cohere → Cohere | Peak RSS (MiB) | 663.3 | Valid | — |
| Cohere → Cohere | Peak RSS high-water mark (MiB) | 663.3 | Valid | — |
| Cohere → Cohere | Time to plateau (s) | 81 | Valid | — |
| Cohere → Cohere | Memory growth (MiB/min) | 0 | Valid | — |
| Cohere → Cohere | Memory load duration (s) | 302 | Valid | — |
| Cohere → Cohere | Memory shape under load | — | Not measured | — |
| Cohere → Cohere | Idle memory shape | — | Not measured | — |
| Cohere → Cohere | Idle memory static | 1 | Valid | — |
| Cohere → Cohere | Idle memory growth (MiB/min) | 0 | Valid | — |
| Cohere → Cohere | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| Cohere → Cohere | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| Cohere → Cohere | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| Cohere → Cohere | Throughput (req/s) | 83 | concurrency 1; n=499 | 50 ms |
| Cohere → Cohere | Throughput (req/s) | 83 | concurrency 1; n=499 | 100 ms |
| Cohere → Cohere | Throughput (req/s) | 112 | concurrency 1,024; n=674 | Unbounded |
| OpenAI → Cohere | Added latency p50 (µs) | 7,551 | Valid · n=790 gateway / 267,951 direct | — |
| OpenAI → Cohere | Added latency p99 (µs) | 8,314 | Valid · n=790 gateway / 267,951 direct | — |
| OpenAI → Cohere | Gateway p99 at concurrency 1 (µs) | 8,341 | Valid · n=790 | — |
| OpenAI → Cohere | Direct-to-mock p99 at concurrency 1 (µs) | 27 | Valid · n=267,951 | — |
| OpenAI → Cohere | CPU per request (µs) | 7,594 | Valid | — |
| OpenAI → Cohere | Requests per CPU-second | 131.7 | Valid | — |
| OpenAI → Cohere | Cost window concurrency | 8 | Valid | — |
| OpenAI → Cohere | Cost window core utilisation | 0.25 | Valid | — |
| OpenAI → Cohere | Cost window successful requests | 798 | Valid | — |
| OpenAI → Cohere | Cost window throughput (req/s) | 132 | Valid | — |
| OpenAI → Cohere | Cost window threads | 13 | Valid | — |
| OpenAI → Cohere | Non-voluntary context switches per request | 0.031 | Valid | — |
| OpenAI → Cohere | Cost window major page faults | 0 | Valid | — |
| OpenAI → Cohere | Added time to first event p50 (µs) | — | Untestable | — |
| OpenAI → Cohere | Added time to first event p99 (µs) | — | Untestable | — |
| OpenAI → Cohere | Added inter-frame gap p50 (µs) | — | Untestable | — |
| OpenAI → Cohere | Added inter-frame gap p99 (µs) | — | Untestable | — |
| OpenAI → Cohere | Streams sustained (frames/s) | — | Untestable | — |
| OpenAI → Cohere | Streams sustained | — | Untestable | — |
| OpenAI → Cohere | Idle RSS (MiB) | 592.6 | Valid | — |
| OpenAI → Cohere | Steady-state RSS (MiB) | 666.1 | Valid | — |
| OpenAI → Cohere | Recovered RSS (MiB) | 666.1 | Valid | — |
| OpenAI → Cohere | Peak RSS (MiB) | 666.1 | Valid | — |
| OpenAI → Cohere | Peak RSS high-water mark (MiB) | 666.1 | Valid | — |
| OpenAI → Cohere | Time to plateau (s) | 61 | Valid | — |
| OpenAI → Cohere | Memory growth (MiB/min) | 0.002 | Valid | — |
| OpenAI → Cohere | Memory load duration (s) | 301 | Valid | — |
| OpenAI → Cohere | Memory shape under load | — | Not measured | — |
| OpenAI → Cohere | Idle memory shape | — | Not measured | — |
| OpenAI → Cohere | Idle memory static | 1 | Valid | — |
| OpenAI → Cohere | Idle memory growth (MiB/min) | 0 | Valid | — |
| OpenAI → Cohere | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| OpenAI → Cohere | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| OpenAI → Cohere | Throughput (req/s) | 131 | concurrency 1; n=787 | 10 ms |
| OpenAI → Cohere | Throughput (req/s) | 131 | concurrency 1; n=787 | 50 ms |
| OpenAI → Cohere | Throughput (req/s) | 131 | concurrency 1; n=787 | 100 ms |
| OpenAI → Cohere | Throughput (req/s) | 138 | concurrency 64; n=829 | Unbounded |
| Bedrock → Cohere | — | — | Not declared: recorded note 1 | — |
| Anthropic → Cohere | Added latency p50 (µs) | 8,248 | Valid · n=723 gateway / 260,556 direct | — |
| Anthropic → Cohere | Added latency p99 (µs) | 9,054 | Valid · n=723 gateway / 260,556 direct | — |
| Anthropic → Cohere | Gateway p99 at concurrency 1 (µs) | 9,082 | Valid · n=723 | — |
| Anthropic → Cohere | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=260,556 | — |
| Anthropic → Cohere | CPU per request (µs) | 8,407.2 | Valid | — |
| Anthropic → Cohere | Requests per CPU-second | 118.9 | Valid | — |
| Anthropic → Cohere | Cost window concurrency | 8 | Valid | — |
| Anthropic → Cohere | Cost window core utilisation | 0.25 | Valid | — |
| Anthropic → Cohere | Cost window successful requests | 722 | Valid | — |
| Anthropic → Cohere | Cost window throughput (req/s) | 120 | Valid | — |
| Anthropic → Cohere | Cost window threads | 13 | Valid | — |
| Anthropic → Cohere | Non-voluntary context switches per request | 0.025 | Valid | — |
| Anthropic → Cohere | Cost window major page faults | 0 | Valid | — |
| Anthropic → Cohere | Added time to first event p50 (µs) | — | Untestable | — |
| Anthropic → Cohere | Added time to first event p99 (µs) | — | Untestable | — |
| Anthropic → Cohere | Added inter-frame gap p50 (µs) | — | Untestable | — |
| Anthropic → Cohere | Added inter-frame gap p99 (µs) | — | Untestable | — |
| Anthropic → Cohere | Streams sustained (frames/s) | — | Untestable | — |
| Anthropic → Cohere | Streams sustained | — | Untestable | — |
| Anthropic → Cohere | Idle RSS (MiB) | 592.2 | Valid | — |
| Anthropic → Cohere | Steady-state RSS (MiB) | 665.8 | Valid | — |
| Anthropic → Cohere | Recovered RSS (MiB) | 665.8 | Valid | — |
| Anthropic → Cohere | Peak RSS (MiB) | 665.8 | Valid | — |
| Anthropic → Cohere | Peak RSS high-water mark (MiB) | 665.8 | Valid | — |
| Anthropic → Cohere | Time to plateau (s) | 61 | Valid | — |
| Anthropic → Cohere | Memory growth (MiB/min) | 0 | Valid | — |
| Anthropic → Cohere | Memory load duration (s) | 302 | Valid | — |
| Anthropic → Cohere | Memory shape under load | — | Not measured | — |
| Anthropic → Cohere | Idle memory shape | — | Not measured | — |
| Anthropic → Cohere | Idle memory static | 1 | Valid | — |
| Anthropic → Cohere | Idle memory growth (MiB/min) | 0 | Valid | — |
| Anthropic → Cohere | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| Anthropic → Cohere | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| Anthropic → Cohere | Throughput (req/s) | 118 | concurrency 1; n=713 | 10 ms |
| Anthropic → Cohere | Throughput (req/s) | 120 | concurrency 2; n=722 | 50 ms |
| Anthropic → Cohere | Throughput (req/s) | 120 | concurrency 2; n=722 | 100 ms |
| Anthropic → Cohere | Throughput (req/s) | 120 | concurrency 2; n=722 | Unbounded |
| OpenAI Responses → Cohere | Added latency p50 (µs) | 9,710 | Valid · n=615 gateway / 263,786 direct | — |
| OpenAI Responses → Cohere | Added latency p99 (µs) | 10,423 | Valid · n=615 gateway / 263,786 direct | — |
| OpenAI Responses → Cohere | Gateway p99 at concurrency 1 (µs) | 10,451 | Valid · n=615 | — |
| OpenAI Responses → Cohere | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=263,786 | — |
| OpenAI Responses → Cohere | CPU per request (µs) | 9,886.4 | Valid | — |
| OpenAI Responses → Cohere | Requests per CPU-second | 101.1 | Valid | — |
| OpenAI Responses → Cohere | Cost window concurrency | 8 | Valid | — |
| OpenAI Responses → Cohere | Cost window core utilisation | 0.251 | Valid | — |
| OpenAI Responses → Cohere | Cost window successful requests | 616 | Valid | — |
| OpenAI Responses → Cohere | Cost window throughput (req/s) | 102 | Valid | — |
| OpenAI Responses → Cohere | Cost window threads | 13 | Valid | — |
| OpenAI Responses → Cohere | Non-voluntary context switches per request | 0.047 | Valid | — |
| OpenAI Responses → Cohere | Cost window major page faults | 0 | Valid | — |
| OpenAI Responses → Cohere | Added time to first event p50 (µs) | — | Untestable | — |
| OpenAI Responses → Cohere | Added time to first event p99 (µs) | — | Untestable | — |
| OpenAI Responses → Cohere | Added inter-frame gap p50 (µs) | — | Untestable | — |
| OpenAI Responses → Cohere | Added inter-frame gap p99 (µs) | — | Untestable | — |
| OpenAI Responses → Cohere | Streams sustained (frames/s) | — | Untestable | — |
| OpenAI Responses → Cohere | Streams sustained | — | Untestable | — |
| OpenAI Responses → Cohere | Idle RSS (MiB) | 592.7 | Valid | — |
| OpenAI Responses → Cohere | Steady-state RSS (MiB) | 667.8 | Valid | — |
| OpenAI Responses → Cohere | Recovered RSS (MiB) | 667.8 | Valid | — |
| OpenAI Responses → Cohere | Peak RSS (MiB) | 667.8 | Valid | — |
| OpenAI Responses → Cohere | Peak RSS high-water mark (MiB) | 667.8 | Valid | — |
| OpenAI Responses → Cohere | Time to plateau (s) | 62 | Valid | — |
| OpenAI Responses → Cohere | Memory growth (MiB/min) | 0 | Valid | — |
| OpenAI Responses → Cohere | Memory load duration (s) | 304 | Valid | — |
| OpenAI Responses → Cohere | Memory shape under load | — | Not measured | — |
| OpenAI Responses → Cohere | Idle memory shape | — | Not measured | — |
| OpenAI Responses → Cohere | Idle memory static | 1 | Valid | — |
| OpenAI Responses → Cohere | Idle memory growth (MiB/min) | 0 | Valid | — |
| OpenAI Responses → Cohere | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| OpenAI Responses → Cohere | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| OpenAI Responses → Cohere | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| OpenAI Responses → Cohere | Throughput (req/s) | 104 | concurrency 2; n=630 | 50 ms |
| OpenAI Responses → Cohere | Throughput (req/s) | 104 | concurrency 2; n=630 | 100 ms |
| OpenAI Responses → Cohere | Throughput (req/s) | 106 | concurrency 64; n=637 | Unbounded |
| Gemini → Cohere | Added latency p50 (µs) | 13,881 | Valid · n=428 gateway / 262,194 direct | — |
| Gemini → Cohere | Added latency p99 (µs) | 14,770 | Valid · n=428 gateway / 262,194 direct | — |
| Gemini → Cohere | Gateway p99 at concurrency 1 (µs) | 14,798 | Valid · n=428 | — |
| Gemini → Cohere | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=262,194 | — |
| Gemini → Cohere | CPU per request (µs) | 14,342.7 | Valid | — |
| Gemini → Cohere | Requests per CPU-second | 69.72 | Valid | — |
| Gemini → Cohere | Cost window concurrency | 8 | Valid | — |
| Gemini → Cohere | Cost window core utilisation | 0.251 | Valid | — |
| Gemini → Cohere | Cost window successful requests | 426 | Valid | — |
| Gemini → Cohere | Cost window throughput (req/s) | 70 | Valid | — |
| Gemini → Cohere | Cost window threads | 13 | Valid | — |
| Gemini → Cohere | Non-voluntary context switches per request | 0.063 | Valid | — |
| Gemini → Cohere | Cost window major page faults | 0 | Valid | — |
| Gemini → Cohere | Added time to first event p50 (µs) | — | Untestable | — |
| Gemini → Cohere | Added time to first event p99 (µs) | — | Untestable | — |
| Gemini → Cohere | Added inter-frame gap p50 (µs) | — | Untestable | — |
| Gemini → Cohere | Added inter-frame gap p99 (µs) | — | Untestable | — |
| Gemini → Cohere | Streams sustained (frames/s) | — | Untestable | — |
| Gemini → Cohere | Streams sustained | — | Untestable | — |
| Gemini → Cohere | Idle RSS (MiB) | 593.4 | Valid | — |
| Gemini → Cohere | Steady-state RSS (MiB) | 670.8 | Valid | — |
| Gemini → Cohere | Recovered RSS (MiB) | 671 | Valid | — |
| Gemini → Cohere | Peak RSS (MiB) | 671 | Valid | — |
| Gemini → Cohere | Peak RSS high-water mark (MiB) | 671 | Valid | — |
| Gemini → Cohere | Time to plateau (s) | 63 | Valid | — |
| Gemini → Cohere | Memory growth (MiB/min) | 0.096 | Valid | — |
| Gemini → Cohere | Memory load duration (s) | 302 | Valid | — |
| Gemini → Cohere | Memory shape under load | — | Not measured | — |
| Gemini → Cohere | Idle memory shape | — | Not measured | — |
| Gemini → Cohere | Idle memory static | 1 | Valid | — |
| Gemini → Cohere | Idle memory growth (MiB/min) | 0 | Valid | — |
| Gemini → Cohere | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| Gemini → Cohere | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| Gemini → Cohere | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| Gemini → Cohere | Throughput (req/s) | 71 | concurrency 1; n=427 | 50 ms |
| Gemini → Cohere | Throughput (req/s) | 71 | concurrency 1; n=427 | 100 ms |
| Gemini → Cohere | Throughput (req/s) | 73 | concurrency 256; n=439 | Unbounded |
| Cohere → Anthropic | — | — | Not declared: recorded note 1 | — |
| Bedrock → Anthropic | — | — | Not declared: recorded note 1 | — |
| OpenAI Responses → Anthropic | Added latency p50 (µs) | 13,061 | Valid · n=457 gateway / 263,938 direct | — |
| OpenAI Responses → Anthropic | Added latency p99 (µs) | 13,826 | Valid · n=457 gateway / 263,938 direct | — |
| OpenAI Responses → Anthropic | Gateway p99 at concurrency 1 (µs) | 13,854 | Valid · n=457 | — |
| OpenAI Responses → Anthropic | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=263,938 | — |
| OpenAI Responses → Anthropic | CPU per request (µs) | 13,414.1 | Valid | — |
| OpenAI Responses → Anthropic | Requests per CPU-second | 74.55 | Valid | — |
| OpenAI Responses → Anthropic | Cost window concurrency | 8 | Valid | — |
| OpenAI Responses → Anthropic | Cost window core utilisation | 0.25 | Valid | — |
| OpenAI Responses → Anthropic | Cost window successful requests | 454 | Valid | — |
| OpenAI Responses → Anthropic | Cost window throughput (req/s) | 75 | Valid | — |
| OpenAI Responses → Anthropic | Cost window threads | 13 | Valid | — |
| OpenAI Responses → Anthropic | Non-voluntary context switches per request | 0.055 | Valid | — |
| OpenAI Responses → Anthropic | Cost window major page faults | 0 | Valid | — |
| OpenAI Responses → Anthropic | Added time to first event p50 (µs) | — | Untestable | — |
| OpenAI Responses → Anthropic | Added time to first event p99 (µs) | — | Untestable | — |
| OpenAI Responses → Anthropic | Added inter-frame gap p50 (µs) | — | Untestable | — |
| OpenAI Responses → Anthropic | Added inter-frame gap p99 (µs) | — | Untestable | — |
| OpenAI Responses → Anthropic | Streams sustained (frames/s) | — | Untestable | — |
| OpenAI Responses → Anthropic | Streams sustained | — | Untestable | — |
| OpenAI Responses → Anthropic | Idle RSS (MiB) | 592.7 | Valid | — |
| OpenAI Responses → Anthropic | Steady-state RSS (MiB) | 667.6 | Valid | — |
| OpenAI Responses → Anthropic | Recovered RSS (MiB) | 667.6 | Valid | — |
| OpenAI Responses → Anthropic | Peak RSS (MiB) | 667.6 | Valid | — |
| OpenAI Responses → Anthropic | Peak RSS high-water mark (MiB) | 667.6 | Valid | — |
| OpenAI Responses → Anthropic | Time to plateau (s) | 62 | Valid | — |
| OpenAI Responses → Anthropic | Memory growth (MiB/min) | 0 | Valid | — |
| OpenAI Responses → Anthropic | Memory load duration (s) | 302 | Valid | — |
| OpenAI Responses → Anthropic | Memory shape under load | — | Not measured | — |
| OpenAI Responses → Anthropic | Idle memory shape | — | Not measured | — |
| OpenAI Responses → Anthropic | Idle memory static | 1 | Valid | — |
| OpenAI Responses → Anthropic | Idle memory growth (MiB/min) | 0 | Valid | — |
| OpenAI Responses → Anthropic | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| OpenAI Responses → Anthropic | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| OpenAI Responses → Anthropic | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| OpenAI Responses → Anthropic | Throughput (req/s) | 76 | concurrency 1; n=458 | 50 ms |
| OpenAI Responses → Anthropic | Throughput (req/s) | 76 | concurrency 1; n=458 | 100 ms |
| OpenAI Responses → Anthropic | Throughput (req/s) | 76 | concurrency 1; n=458 | Unbounded |
| Gemini → Anthropic | Added latency p50 (µs) | 16,324 | Valid · n=364 gateway / 262,033 direct | — |
| Gemini → Anthropic | Added latency p99 (µs) | 17,226 | Valid · n=364 gateway / 262,033 direct | — |
| Gemini → Anthropic | Gateway p99 at concurrency 1 (µs) | 17,254 | Valid · n=364 | — |
| Gemini → Anthropic | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=262,033 | — |
| Gemini → Anthropic | CPU per request (µs) | 16,832 | Valid | — |
| Gemini → Anthropic | Requests per CPU-second | 59.41 | Valid | — |
| Gemini → Anthropic | Cost window concurrency | 8 | Valid | — |
| Gemini → Anthropic | Cost window core utilisation | 0.25 | Valid | — |
| Gemini → Anthropic | Cost window successful requests | 363 | Valid | — |
| Gemini → Anthropic | Cost window throughput (req/s) | 60 | Valid | — |
| Gemini → Anthropic | Cost window threads | 13 | Valid | — |
| Gemini → Anthropic | Non-voluntary context switches per request | 0.074 | Valid | — |
| Gemini → Anthropic | Cost window major page faults | 0 | Valid | — |
| Gemini → Anthropic | Added time to first event p50 (µs) | — | Untestable | — |
| Gemini → Anthropic | Added time to first event p99 (µs) | — | Untestable | — |
| Gemini → Anthropic | Added inter-frame gap p50 (µs) | — | Untestable | — |
| Gemini → Anthropic | Added inter-frame gap p99 (µs) | — | Untestable | — |
| Gemini → Anthropic | Streams sustained (frames/s) | — | Untestable | — |
| Gemini → Anthropic | Streams sustained | — | Untestable | — |
| Gemini → Anthropic | Idle RSS (MiB) | 591.9 | Valid | — |
| Gemini → Anthropic | Steady-state RSS (MiB) | 671.9 | Valid | — |
| Gemini → Anthropic | Recovered RSS (MiB) | 671.9 | Valid | — |
| Gemini → Anthropic | Peak RSS (MiB) | 671.9 | Valid | — |
| Gemini → Anthropic | Peak RSS high-water mark (MiB) | 671.9 | Valid | — |
| Gemini → Anthropic | Time to plateau (s) | 63 | Valid | — |
| Gemini → Anthropic | Memory growth (MiB/min) | 0 | Valid | — |
| Gemini → Anthropic | Memory load duration (s) | 305 | Valid | — |
| Gemini → Anthropic | Memory shape under load | — | Not measured | — |
| Gemini → Anthropic | Idle memory shape | — | Not measured | — |
| Gemini → Anthropic | Idle memory static | 1 | Valid | — |
| Gemini → Anthropic | Idle memory growth (MiB/min) | 0 | Valid | — |
| Gemini → Anthropic | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| Gemini → Anthropic | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| Gemini → Anthropic | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| Gemini → Anthropic | Throughput (req/s) | 60 | concurrency 2; n=362 | 50 ms |
| Gemini → Anthropic | Throughput (req/s) | 60 | concurrency 2; n=362 | 100 ms |
| Gemini → Anthropic | Throughput (req/s) | 60 | concurrency 2; n=362 | Unbounded |
| OpenAI → Anthropic | Added latency p50 (µs) | 10,313 | Valid · n=579 gateway / 265,135 direct | — |
| OpenAI → Anthropic | Added latency p99 (µs) | 11,125 | Valid · n=579 gateway / 265,135 direct | — |
| OpenAI → Anthropic | Gateway p99 at concurrency 1 (µs) | 11,153 | Valid · n=579 | — |
| OpenAI → Anthropic | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=265,135 | — |
| OpenAI → Anthropic | CPU per request (µs) | 10,411.7 | Valid | — |
| OpenAI → Anthropic | Requests per CPU-second | 96.05 | Valid | — |
| OpenAI → Anthropic | Cost window concurrency | 8 | Valid | — |
| OpenAI → Anthropic | Cost window core utilisation | 0.25 | Valid | — |
| OpenAI → Anthropic | Cost window successful requests | 583 | Valid | — |
| OpenAI → Anthropic | Cost window throughput (req/s) | 97 | Valid | — |
| OpenAI → Anthropic | Cost window threads | 28 | Valid | — |
| OpenAI → Anthropic | Non-voluntary context switches per request | 0.038 | Valid | — |
| OpenAI → Anthropic | Cost window major page faults | 0 | Valid | — |
| OpenAI → Anthropic | Added time to first event p50 (µs) | 13,428 | Valid · n=100 gateway / 100 direct | — |
| OpenAI → Anthropic | Added time to first event p99 (µs) | 14,151 | Valid · n=100 gateway / 100 direct | — |
| OpenAI → Anthropic | Added inter-frame gap p50 (µs) | 3 | Valid · n=63 gateway / 63 direct | — |
| OpenAI → Anthropic | Added inter-frame gap p99 (µs) | 1,348 | Valid · n=63 gateway / 63 direct | — |
| OpenAI → Anthropic | Streams sustained (frames/s) | 1,232.9 | Valid | — |
| OpenAI → Anthropic | Streams sustained | 117 | Valid | — |
| OpenAI → Anthropic | Idle RSS (MiB) | 592.5 | Valid | — |
| OpenAI → Anthropic | Steady-state RSS (MiB) | 667.1 | Valid | — |
| OpenAI → Anthropic | Recovered RSS (MiB) | 667.1 | Valid | — |
| OpenAI → Anthropic | Peak RSS (MiB) | 667.1 | Valid | — |
| OpenAI → Anthropic | Peak RSS high-water mark (MiB) | 667.1 | Valid | — |
| OpenAI → Anthropic | Time to plateau (s) | 62 | Valid | — |
| OpenAI → Anthropic | Memory growth (MiB/min) | 0 | Valid | — |
| OpenAI → Anthropic | Memory load duration (s) | 304 | Valid | — |
| OpenAI → Anthropic | Memory shape under load | — | Not measured | — |
| OpenAI → Anthropic | Idle memory shape | — | Not measured | — |
| OpenAI → Anthropic | Idle memory static | 1 | Valid | — |
| OpenAI → Anthropic | Idle memory growth (MiB/min) | 0 | Valid | — |
| OpenAI → Anthropic | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| OpenAI → Anthropic | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| OpenAI → Anthropic | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| OpenAI → Anthropic | Throughput (req/s) | 97 | concurrency 1; n=586 | 50 ms |
| OpenAI → Anthropic | Throughput (req/s) | 98 | concurrency 8; n=589 | 100 ms |
| OpenAI → Anthropic | Throughput (req/s) | 98 | concurrency 8; n=589 | Unbounded |
| Anthropic → Anthropic | Added latency p50 (µs) | 10,994 | Valid · n=544 gateway / 263,642 direct | — |
| Anthropic → Anthropic | Added latency p99 (µs) | 11,745 | Valid · n=544 gateway / 263,642 direct | — |
| Anthropic → Anthropic | Gateway p99 at concurrency 1 (µs) | 11,773 | Valid · n=544 | — |
| Anthropic → Anthropic | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=263,642 | — |
| Anthropic → Anthropic | CPU per request (µs) | 11,322.2 | Valid | — |
| Anthropic → Anthropic | Requests per CPU-second | 88.32 | Valid | — |
| Anthropic → Anthropic | Cost window concurrency | 8 | Valid | — |
| Anthropic → Anthropic | Cost window core utilisation | 0.25 | Valid | — |
| Anthropic → Anthropic | Cost window successful requests | 537 | Valid | — |
| Anthropic → Anthropic | Cost window throughput (req/s) | 89 | Valid | — |
| Anthropic → Anthropic | Cost window threads | 268 | Valid | — |
| Anthropic → Anthropic | Non-voluntary context switches per request | 0.039 | Valid | — |
| Anthropic → Anthropic | Cost window major page faults | 0 | Valid | — |
| Anthropic → Anthropic | Added time to first event p50 (µs) | 14,510 | Valid · n=100 gateway / 100 direct | — |
| Anthropic → Anthropic | Added time to first event p99 (µs) | 15,815 | Valid · n=100 gateway / 100 direct | — |
| Anthropic → Anthropic | Added inter-frame gap p50 (µs) | 0 | Valid · n=63 gateway / 63 direct | — |
| Anthropic → Anthropic | Added inter-frame gap p99 (µs) | — | Below resolution · n=63 gateway / 63 direct | — |
| Anthropic → Anthropic | Streams sustained (frames/s) | 3,187.6 | Valid | — |
| Anthropic → Anthropic | Streams sustained | 255 | Valid | — |
| Anthropic → Anthropic | Idle RSS (MiB) | 592.5 | Valid | — |
| Anthropic → Anthropic | Steady-state RSS (MiB) | 658.1 | Valid | — |
| Anthropic → Anthropic | Recovered RSS (MiB) | 658.1 | Valid | — |
| Anthropic → Anthropic | Peak RSS (MiB) | 658.1 | Valid | — |
| Anthropic → Anthropic | Peak RSS high-water mark (MiB) | 658.1 | Valid | — |
| Anthropic → Anthropic | Time to plateau (s) | 62 | Valid | — |
| Anthropic → Anthropic | Memory growth (MiB/min) | 0 | Valid | — |
| Anthropic → Anthropic | Memory load duration (s) | 300 | Valid | — |
| Anthropic → Anthropic | Memory shape under load | — | Not measured | — |
| Anthropic → Anthropic | Idle memory shape | — | Not measured | — |
| Anthropic → Anthropic | Idle memory static | 1 | Valid | — |
| Anthropic → Anthropic | Idle memory growth (MiB/min) | 0 | Valid | — |
| Anthropic → Anthropic | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| Anthropic → Anthropic | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| Anthropic → Anthropic | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| Anthropic → Anthropic | Throughput (req/s) | 89 | concurrency 2; n=538 | 50 ms |
| Anthropic → Anthropic | Throughput (req/s) | 89 | concurrency 2; n=538 | 100 ms |
| Anthropic → Anthropic | Throughput (req/s) | 91 | concurrency 64; n=550 | Unbounded |
| OpenAI Responses → Bedrock | Added latency p50 (µs) | 17,646 | Valid · n=339 gateway / 267,586 direct | — |
| OpenAI Responses → Bedrock | Added latency p99 (µs) | 18,361 | Valid · n=339 gateway / 267,586 direct | — |
| OpenAI Responses → Bedrock | Gateway p99 at concurrency 1 (µs) | 18,388 | Valid · n=339 | — |
| OpenAI Responses → Bedrock | Direct-to-mock p99 at concurrency 1 (µs) | 27 | Valid · n=267,586 | — |
| OpenAI Responses → Bedrock | CPU per request (µs) | 18,244 | Valid | — |
| OpenAI Responses → Bedrock | Requests per CPU-second | 54.81 | Valid | — |
| OpenAI Responses → Bedrock | Cost window concurrency | 8 | Valid | — |
| OpenAI Responses → Bedrock | Cost window core utilisation | 0.25 | Valid | — |
| OpenAI Responses → Bedrock | Cost window successful requests | 336 | Valid | — |
| OpenAI Responses → Bedrock | Cost window throughput (req/s) | 55 | Valid | — |
| OpenAI Responses → Bedrock | Cost window threads | 17 | Valid | — |
| OpenAI Responses → Bedrock | Non-voluntary context switches per request | 0.074 | Valid | — |
| OpenAI Responses → Bedrock | Cost window major page faults | 0 | Valid | — |
| OpenAI Responses → Bedrock | Added time to first event p50 (µs) | — | Untestable | — |
| OpenAI Responses → Bedrock | Added time to first event p99 (µs) | — | Untestable | — |
| OpenAI Responses → Bedrock | Added inter-frame gap p50 (µs) | — | Untestable | — |
| OpenAI Responses → Bedrock | Added inter-frame gap p99 (µs) | — | Untestable | — |
| OpenAI Responses → Bedrock | Streams sustained (frames/s) | — | Untestable | — |
| OpenAI Responses → Bedrock | Streams sustained | — | Untestable | — |
| OpenAI Responses → Bedrock | Idle RSS (MiB) | 593 | Valid | — |
| OpenAI Responses → Bedrock | Steady-state RSS (MiB) | 689.4 | Valid | — |
| OpenAI Responses → Bedrock | Recovered RSS (MiB) | 689.4 | Valid | — |
| OpenAI Responses → Bedrock | Peak RSS (MiB) | 689.4 | Valid | — |
| OpenAI Responses → Bedrock | Peak RSS high-water mark (MiB) | 689.4 | Valid | — |
| OpenAI Responses → Bedrock | Time to plateau (s) | 64 | Valid | — |
| OpenAI Responses → Bedrock | Memory growth (MiB/min) | 0 | Valid | — |
| OpenAI Responses → Bedrock | Memory load duration (s) | 301 | Valid | — |
| OpenAI Responses → Bedrock | Memory shape under load | — | Not measured | — |
| OpenAI Responses → Bedrock | Idle memory shape | — | Not measured | — |
| OpenAI Responses → Bedrock | Idle memory static | 1 | Valid | — |
| OpenAI Responses → Bedrock | Idle memory growth (MiB/min) | 0 | Valid | — |
| OpenAI Responses → Bedrock | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| OpenAI Responses → Bedrock | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| OpenAI Responses → Bedrock | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| OpenAI Responses → Bedrock | Throughput (req/s) | 56 | concurrency 1; n=338 | 50 ms |
| OpenAI Responses → Bedrock | Throughput (req/s) | 56 | concurrency 1; n=338 | 100 ms |
| OpenAI Responses → Bedrock | Throughput (req/s) | 57 | concurrency 2,048; n=345 | Unbounded |
| Cohere → Bedrock | — | — | Not declared: recorded note 1 | — |
| Anthropic → Bedrock | Added latency p50 (µs) | 16,315 | Valid · n=366 gateway / 262,394 direct | — |
| Anthropic → Bedrock | Added latency p99 (µs) | 17,156 | Valid · n=366 gateway / 262,394 direct | — |
| Anthropic → Bedrock | Gateway p99 at concurrency 1 (µs) | 17,184 | Valid · n=366 | — |
| Anthropic → Bedrock | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=262,394 | — |
| Anthropic → Bedrock | CPU per request (µs) | 18,047.3 | Valid | — |
| Anthropic → Bedrock | Requests per CPU-second | 55.41 | Valid | — |
| Anthropic → Bedrock | Cost window concurrency | 8 | Valid | — |
| Anthropic → Bedrock | Cost window core utilisation | 0.25 | Valid | — |
| Anthropic → Bedrock | Cost window successful requests | 338 | Valid | — |
| Anthropic → Bedrock | Cost window throughput (req/s) | 56 | Valid | — |
| Anthropic → Bedrock | Cost window threads | 17 | Valid | — |
| Anthropic → Bedrock | Non-voluntary context switches per request | 0.074 | Valid | — |
| Anthropic → Bedrock | Cost window major page faults | 0 | Valid | — |
| Anthropic → Bedrock | Added time to first event p50 (µs) | — | Untestable | — |
| Anthropic → Bedrock | Added time to first event p99 (µs) | — | Untestable | — |
| Anthropic → Bedrock | Added inter-frame gap p50 (µs) | — | Untestable | — |
| Anthropic → Bedrock | Added inter-frame gap p99 (µs) | — | Untestable | — |
| Anthropic → Bedrock | Streams sustained (frames/s) | — | Untestable | — |
| Anthropic → Bedrock | Streams sustained | — | Untestable | — |
| Anthropic → Bedrock | Idle RSS (MiB) | 592.7 | Valid | — |
| Anthropic → Bedrock | Steady-state RSS (MiB) | 687.5 | Valid | — |
| Anthropic → Bedrock | Recovered RSS (MiB) | 687.5 | Valid | — |
| Anthropic → Bedrock | Peak RSS (MiB) | 687.5 | Valid | — |
| Anthropic → Bedrock | Peak RSS high-water mark (MiB) | 687.5 | Valid | — |
| Anthropic → Bedrock | Time to plateau (s) | 63 | Valid | — |
| Anthropic → Bedrock | Memory growth (MiB/min) | 0 | Valid | — |
| Anthropic → Bedrock | Memory load duration (s) | 305 | Valid | — |
| Anthropic → Bedrock | Memory shape under load | — | Not measured | — |
| Anthropic → Bedrock | Idle memory shape | — | Not measured | — |
| Anthropic → Bedrock | Idle memory static | 1 | Valid | — |
| Anthropic → Bedrock | Idle memory growth (MiB/min) | 0 | Valid | — |
| Anthropic → Bedrock | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| Anthropic → Bedrock | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| Anthropic → Bedrock | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| Anthropic → Bedrock | Throughput (req/s) | 62 | concurrency 2; n=376 | 50 ms |
| Anthropic → Bedrock | Throughput (req/s) | 62 | concurrency 2; n=376 | 100 ms |
| Anthropic → Bedrock | Throughput (req/s) | 63 | concurrency 64; n=382 | Unbounded |
| Gemini → Bedrock | Added latency p50 (µs) | 21,239 | Valid · n=281 gateway / 267,100 direct | — |
| Gemini → Bedrock | Added latency p99 (µs) | 22,150 | Valid · n=281 gateway / 267,100 direct | — |
| Gemini → Bedrock | Gateway p99 at concurrency 1 (µs) | 22,177 | Valid · n=281 | — |
| Gemini → Bedrock | Direct-to-mock p99 at concurrency 1 (µs) | 27 | Valid · n=267,100 | — |
| Gemini → Bedrock | CPU per request (µs) | 22,086.3 | Valid | — |
| Gemini → Bedrock | Requests per CPU-second | 45.28 | Valid | — |
| Gemini → Bedrock | Cost window concurrency | 8 | Valid | — |
| Gemini → Bedrock | Cost window core utilisation | 0.251 | Valid | — |
| Gemini → Bedrock | Cost window successful requests | 278 | Valid | — |
| Gemini → Bedrock | Cost window throughput (req/s) | 46 | Valid | — |
| Gemini → Bedrock | Cost window threads | 17 | Valid | — |
| Gemini → Bedrock | Non-voluntary context switches per request | 0.079 | Valid | — |
| Gemini → Bedrock | Cost window major page faults | 0 | Valid | — |
| Gemini → Bedrock | Added time to first event p50 (µs) | — | Untestable | — |
| Gemini → Bedrock | Added time to first event p99 (µs) | — | Untestable | — |
| Gemini → Bedrock | Added inter-frame gap p50 (µs) | — | Untestable | — |
| Gemini → Bedrock | Added inter-frame gap p99 (µs) | — | Untestable | — |
| Gemini → Bedrock | Streams sustained (frames/s) | — | Untestable | — |
| Gemini → Bedrock | Streams sustained | — | Untestable | — |
| Gemini → Bedrock | Idle RSS (MiB) | 591.9 | Valid | — |
| Gemini → Bedrock | Steady-state RSS (MiB) | 692.8 | Valid | — |
| Gemini → Bedrock | Recovered RSS (MiB) | 692.8 | Valid | — |
| Gemini → Bedrock | Peak RSS (MiB) | 692.8 | Valid | — |
| Gemini → Bedrock | Peak RSS high-water mark (MiB) | 692.8 | Valid | — |
| Gemini → Bedrock | Time to plateau (s) | 64 | Valid | — |
| Gemini → Bedrock | Memory growth (MiB/min) | 0.005 | Valid | — |
| Gemini → Bedrock | Memory load duration (s) | 302 | Valid | — |
| Gemini → Bedrock | Memory shape under load | — | Not measured | — |
| Gemini → Bedrock | Idle memory shape | — | Not measured | — |
| Gemini → Bedrock | Idle memory static | 1 | Valid | — |
| Gemini → Bedrock | Idle memory growth (MiB/min) | 0 | Valid | — |
| Gemini → Bedrock | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| Gemini → Bedrock | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| Gemini → Bedrock | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| Gemini → Bedrock | Throughput (req/s) | 46 | concurrency 1; n=280 | 50 ms |
| Gemini → Bedrock | Throughput (req/s) | 46 | concurrency 1; n=280 | 100 ms |
| Gemini → Bedrock | Throughput (req/s) | 46 | concurrency 1; n=280 | Unbounded |
| Bedrock → Bedrock | Added latency p50 (µs) | 10,032 | Valid · n=595 gateway / 267,885 direct | — |
| Bedrock → Bedrock | Added latency p99 (µs) | 10,847 | Valid · n=595 gateway / 267,885 direct | — |
| Bedrock → Bedrock | Gateway p99 at concurrency 1 (µs) | 10,874 | Valid · n=595 | — |
| Bedrock → Bedrock | Direct-to-mock p99 at concurrency 1 (µs) | 27 | Valid · n=267,885 | — |
| Bedrock → Bedrock | CPU per request (µs) | 10,134 | Valid | — |
| Bedrock → Bedrock | Requests per CPU-second | 98.68 | Valid | — |
| Bedrock → Bedrock | Cost window concurrency | 8 | Valid | — |
| Bedrock → Bedrock | Cost window core utilisation | 0.25 | Valid | — |
| Bedrock → Bedrock | Cost window successful requests | 597 | Valid | — |
| Bedrock → Bedrock | Cost window throughput (req/s) | 99 | Valid | — |
| Bedrock → Bedrock | Cost window threads | 13 | Valid | — |
| Bedrock → Bedrock | Non-voluntary context switches per request | 0.045 | Valid | — |
| Bedrock → Bedrock | Cost window major page faults | 0 | Valid | — |
| Bedrock → Bedrock | Added time to first event p50 (µs) | — | Untestable | — |
| Bedrock → Bedrock | Added time to first event p99 (µs) | — | Untestable | — |
| Bedrock → Bedrock | Added inter-frame gap p50 (µs) | — | Untestable | — |
| Bedrock → Bedrock | Added inter-frame gap p99 (µs) | — | Untestable | — |
| Bedrock → Bedrock | Streams sustained (frames/s) | — | Untestable | — |
| Bedrock → Bedrock | Streams sustained | — | Untestable | — |
| Bedrock → Bedrock | Idle RSS (MiB) | 592.5 | Valid | — |
| Bedrock → Bedrock | Steady-state RSS (MiB) | 671.4 | Valid | — |
| Bedrock → Bedrock | Recovered RSS (MiB) | 671.4 | Valid | — |
| Bedrock → Bedrock | Peak RSS (MiB) | 671.4 | Valid | — |
| Bedrock → Bedrock | Peak RSS high-water mark (MiB) | 671.4 | Valid | — |
| Bedrock → Bedrock | Time to plateau (s) | 61 | Valid | — |
| Bedrock → Bedrock | Memory growth (MiB/min) | 0 | Valid | — |
| Bedrock → Bedrock | Memory load duration (s) | 303 | Valid | — |
| Bedrock → Bedrock | Memory shape under load | — | Not measured | — |
| Bedrock → Bedrock | Idle memory shape | — | Not measured | — |
| Bedrock → Bedrock | Idle memory static | 1 | Valid | — |
| Bedrock → Bedrock | Idle memory growth (MiB/min) | 0 | Valid | — |
| Bedrock → Bedrock | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| Bedrock → Bedrock | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| Bedrock → Bedrock | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| Bedrock → Bedrock | Throughput (req/s) | 101 | concurrency 2; n=607 | 50 ms |
| Bedrock → Bedrock | Throughput (req/s) | 101 | concurrency 2; n=607 | 100 ms |
| Bedrock → Bedrock | Throughput (req/s) | 104 | concurrency 128; n=625 | Unbounded |
| OpenAI → Bedrock | Added latency p50 (µs) | 14,005 | Valid · n=427 gateway / 272,464 direct | — |
| OpenAI → Bedrock | Added latency p99 (µs) | 14,745 | Valid · n=427 gateway / 272,464 direct | — |
| OpenAI → Bedrock | Gateway p99 at concurrency 1 (µs) | 14,772 | Valid · n=427 | — |
| OpenAI → Bedrock | Direct-to-mock p99 at concurrency 1 (µs) | 27 | Valid · n=272,464 | — |
| OpenAI → Bedrock | CPU per request (µs) | 14,262.3 | Valid | — |
| OpenAI → Bedrock | Requests per CPU-second | 70.11 | Valid | — |
| OpenAI → Bedrock | Cost window concurrency | 8 | Valid | — |
| OpenAI → Bedrock | Cost window core utilisation | 0.25 | Valid | — |
| OpenAI → Bedrock | Cost window successful requests | 427 | Valid | — |
| OpenAI → Bedrock | Cost window throughput (req/s) | 71 | Valid | — |
| OpenAI → Bedrock | Cost window threads | 17 | Valid | — |
| OpenAI → Bedrock | Non-voluntary context switches per request | 0.047 | Valid | — |
| OpenAI → Bedrock | Cost window major page faults | 0 | Valid | — |
| OpenAI → Bedrock | Added time to first event p50 (µs) | — | Untestable | — |
| OpenAI → Bedrock | Added time to first event p99 (µs) | — | Untestable | — |
| OpenAI → Bedrock | Added inter-frame gap p50 (µs) | — | Untestable | — |
| OpenAI → Bedrock | Added inter-frame gap p99 (µs) | — | Untestable | — |
| OpenAI → Bedrock | Streams sustained (frames/s) | — | Untestable | — |
| OpenAI → Bedrock | Streams sustained | — | Untestable | — |
| OpenAI → Bedrock | Idle RSS (MiB) | 593.2 | Valid | — |
| OpenAI → Bedrock | Steady-state RSS (MiB) | 689.1 | Valid | — |
| OpenAI → Bedrock | Recovered RSS (MiB) | 689.1 | Valid | — |
| OpenAI → Bedrock | Peak RSS (MiB) | 689.1 | Valid | — |
| OpenAI → Bedrock | Peak RSS high-water mark (MiB) | 689.1 | Valid | — |
| OpenAI → Bedrock | Time to plateau (s) | 63 | Valid | — |
| OpenAI → Bedrock | Memory growth (MiB/min) | 0 | Valid | — |
| OpenAI → Bedrock | Memory load duration (s) | 302 | Valid | — |
| OpenAI → Bedrock | Memory shape under load | — | Not measured | — |
| OpenAI → Bedrock | Idle memory shape | — | Not measured | — |
| OpenAI → Bedrock | Idle memory static | 1 | Valid | — |
| OpenAI → Bedrock | Idle memory growth (MiB/min) | 0 | Valid | — |
| OpenAI → Bedrock | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| OpenAI → Bedrock | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| OpenAI → Bedrock | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| OpenAI → Bedrock | Throughput (req/s) | 72 | concurrency 2; n=434 | 50 ms |
| OpenAI → Bedrock | Throughput (req/s) | 72 | concurrency 2; n=434 | 100 ms |
| OpenAI → Bedrock | Throughput (req/s) | 72 | concurrency 2; n=434 | Unbounded |
| OpenAI → OpenAI | Added latency p50 (µs) | 8,512 | Valid · n=701 gateway / 267,297 direct | — |
| OpenAI → OpenAI | Added latency p99 (µs) | 9,522 | Valid · n=701 gateway / 267,297 direct | — |
| OpenAI → OpenAI | Gateway p99 at concurrency 1 (µs) | 9,549 | Valid · n=701 | — |
| OpenAI → OpenAI | Direct-to-mock p99 at concurrency 1 (µs) | 27 | Valid · n=267,297 | — |
| OpenAI → OpenAI | CPU per request (µs) | 8,644.8 | Valid | — |
| OpenAI → OpenAI | Requests per CPU-second | 115.7 | Valid | — |
| OpenAI → OpenAI | Cost window concurrency | 8 | Valid | — |
| OpenAI → OpenAI | Cost window core utilisation | 0.25 | Valid | — |
| OpenAI → OpenAI | Cost window successful requests | 701 | Valid | — |
| OpenAI → OpenAI | Cost window throughput (req/s) | 116 | Valid | — |
| OpenAI → OpenAI | Cost window threads | 32 | Valid | — |
| OpenAI → OpenAI | Non-voluntary context switches per request | 0.029 | Valid | — |
| OpenAI → OpenAI | Cost window major page faults | 0 | Valid | — |
| OpenAI → OpenAI | Added time to first event p50 (µs) | 11,154 | Valid · n=100 gateway / 100 direct | — |
| OpenAI → OpenAI | Added time to first event p99 (µs) | 11,974 | Valid · n=100 gateway / 100 direct | — |
| OpenAI → OpenAI | Added inter-frame gap p50 (µs) | — | Below resolution · n=63 gateway / 63 direct | — |
| OpenAI → OpenAI | Added inter-frame gap p99 (µs) | 9 | Valid · n=63 gateway / 63 direct | — |
| OpenAI → OpenAI | Streams sustained (frames/s) | 1,208 | Valid | — |
| OpenAI → OpenAI | Streams sustained | 62 | Valid | — |
| OpenAI → OpenAI | Idle RSS (MiB) | 593 | Valid | — |
| OpenAI → OpenAI | Steady-state RSS (MiB) | 628.6 | Valid | — |
| OpenAI → OpenAI | Recovered RSS (MiB) | 628.6 | Valid | — |
| OpenAI → OpenAI | Peak RSS (MiB) | 628.6 | Valid | — |
| OpenAI → OpenAI | Peak RSS high-water mark (MiB) | 628.6 | Valid | — |
| OpenAI → OpenAI | Time to plateau (s) | 61 | Valid | — |
| OpenAI → OpenAI | Memory growth (MiB/min) | 0 | Valid | — |
| OpenAI → OpenAI | Memory load duration (s) | 302 | Valid | — |
| OpenAI → OpenAI | Memory shape under load | — | Not measured | — |
| OpenAI → OpenAI | Idle memory shape | — | Not measured | — |
| OpenAI → OpenAI | Idle memory static | 1 | Valid | — |
| OpenAI → OpenAI | Idle memory growth (MiB/min) | 0 | Valid | — |
| OpenAI → OpenAI | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| OpenAI → OpenAI | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| OpenAI → OpenAI | Throughput (req/s) | 117 | concurrency 1; n=708 | 10 ms |
| OpenAI → OpenAI | Throughput (req/s) | 119 | concurrency 2; n=718 | 50 ms |
| OpenAI → OpenAI | Throughput (req/s) | 119 | concurrency 2; n=718 | 100 ms |
| OpenAI → OpenAI | Throughput (req/s) | 119 | concurrency 2; n=718 | Unbounded |
| Anthropic → OpenAI | — | — | Not served: probed and answered HTTP 200, but the body is not a usable anthropic response (see body_snippet) | — |
| OpenAI Responses → OpenAI | Added latency p50 (µs) | 10,676 | Valid · n=559 gateway / 265,389 direct | — |
| OpenAI Responses → OpenAI | Added latency p99 (µs) | 11,606 | Valid · n=559 gateway / 265,389 direct | — |
| OpenAI Responses → OpenAI | Gateway p99 at concurrency 1 (µs) | 11,634 | Valid · n=559 | — |
| OpenAI Responses → OpenAI | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=265,389 | — |
| OpenAI Responses → OpenAI | CPU per request (µs) | 10,555.6 | Valid | — |
| OpenAI Responses → OpenAI | Requests per CPU-second | 94.74 | Valid | — |
| OpenAI Responses → OpenAI | Cost window concurrency | 8 | Valid | — |
| OpenAI Responses → OpenAI | Cost window core utilisation | 0.25 | Valid | — |
| OpenAI Responses → OpenAI | Cost window successful requests | 576 | Valid | — |
| OpenAI Responses → OpenAI | Cost window throughput (req/s) | 95 | Valid | — |
| OpenAI Responses → OpenAI | Cost window threads | 13 | Valid | — |
| OpenAI Responses → OpenAI | Non-voluntary context switches per request | 0.059 | Valid | — |
| OpenAI Responses → OpenAI | Cost window major page faults | 0 | Valid | — |
| OpenAI Responses → OpenAI | Added time to first event p50 (µs) | — | Untestable | — |
| OpenAI Responses → OpenAI | Added time to first event p99 (µs) | — | Untestable | — |
| OpenAI Responses → OpenAI | Added inter-frame gap p50 (µs) | — | Untestable | — |
| OpenAI Responses → OpenAI | Added inter-frame gap p99 (µs) | — | Untestable | — |
| OpenAI Responses → OpenAI | Streams sustained (frames/s) | — | Untestable | — |
| OpenAI Responses → OpenAI | Streams sustained | — | Untestable | — |
| OpenAI Responses → OpenAI | Idle RSS (MiB) | 591.9 | Valid | — |
| OpenAI Responses → OpenAI | Steady-state RSS (MiB) | 630 | Valid | — |
| OpenAI Responses → OpenAI | Recovered RSS (MiB) | 630 | Valid | — |
| OpenAI Responses → OpenAI | Peak RSS (MiB) | 630 | Valid | — |
| OpenAI Responses → OpenAI | Peak RSS high-water mark (MiB) | 630 | Valid | — |
| OpenAI Responses → OpenAI | Time to plateau (s) | 62 | Valid | — |
| OpenAI Responses → OpenAI | Memory growth (MiB/min) | 0.006 | Valid | — |
| OpenAI Responses → OpenAI | Memory load duration (s) | 304 | Valid | — |
| OpenAI Responses → OpenAI | Memory shape under load | — | Not measured | — |
| OpenAI Responses → OpenAI | Idle memory shape | — | Not measured | — |
| OpenAI Responses → OpenAI | Idle memory static | 1 | Valid | — |
| OpenAI Responses → OpenAI | Idle memory growth (MiB/min) | 0 | Valid | — |
| OpenAI Responses → OpenAI | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| OpenAI Responses → OpenAI | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| OpenAI Responses → OpenAI | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| OpenAI Responses → OpenAI | Throughput (req/s) | 95 | concurrency 2; n=572 | 50 ms |
| OpenAI Responses → OpenAI | Throughput (req/s) | 95 | concurrency 2; n=572 | 100 ms |
| OpenAI Responses → OpenAI | Throughput (req/s) | 95 | concurrency 2; n=572 | Unbounded |
| Cohere → OpenAI | — | — | Not declared: recorded note 1 | — |
| Bedrock → OpenAI | — | — | Not declared: recorded note 1 | — |
| Gemini → OpenAI | Added latency p50 (µs) | 14,407 | Valid · n=412 gateway / 263,451 direct | — |
| Gemini → OpenAI | Added latency p99 (µs) | 15,293 | Valid · n=412 gateway / 263,451 direct | — |
| Gemini → OpenAI | Gateway p99 at concurrency 1 (µs) | 15,321 | Valid · n=412 | — |
| Gemini → OpenAI | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=263,451 | — |
| Gemini → OpenAI | CPU per request (µs) | 15,086.4 | Valid | — |
| Gemini → OpenAI | Requests per CPU-second | 66.28 | Valid | — |
| Gemini → OpenAI | Cost window concurrency | 8 | Valid | — |
| Gemini → OpenAI | Cost window core utilisation | 0.25 | Valid | — |
| Gemini → OpenAI | Cost window successful requests | 405 | Valid | — |
| Gemini → OpenAI | Cost window throughput (req/s) | 67 | Valid | — |
| Gemini → OpenAI | Cost window threads | 13 | Valid | — |
| Gemini → OpenAI | Non-voluntary context switches per request | 0.044 | Valid | — |
| Gemini → OpenAI | Cost window major page faults | 0 | Valid | — |
| Gemini → OpenAI | Added time to first event p50 (µs) | — | Untestable | — |
| Gemini → OpenAI | Added time to first event p99 (µs) | — | Untestable | — |
| Gemini → OpenAI | Added inter-frame gap p50 (µs) | — | Untestable | — |
| Gemini → OpenAI | Added inter-frame gap p99 (µs) | — | Untestable | — |
| Gemini → OpenAI | Streams sustained (frames/s) | — | Untestable | — |
| Gemini → OpenAI | Streams sustained | — | Untestable | — |
| Gemini → OpenAI | Idle RSS (MiB) | 592 | Valid | — |
| Gemini → OpenAI | Steady-state RSS (MiB) | 632.3 | Valid | — |
| Gemini → OpenAI | Recovered RSS (MiB) | 632.3 | Valid | — |
| Gemini → OpenAI | Peak RSS (MiB) | 632.3 | Valid | — |
| Gemini → OpenAI | Peak RSS high-water mark (MiB) | 632.3 | Valid | — |
| Gemini → OpenAI | Time to plateau (s) | 63 | Valid | — |
| Gemini → OpenAI | Memory growth (MiB/min) | 0.003 | Valid | — |
| Gemini → OpenAI | Memory load duration (s) | 302 | Valid | — |
| Gemini → OpenAI | Memory shape under load | — | Not measured | — |
| Gemini → OpenAI | Idle memory shape | — | Not measured | — |
| Gemini → OpenAI | Idle memory static | 1 | Valid | — |
| Gemini → OpenAI | Idle memory growth (MiB/min) | 0 | Valid | — |
| Gemini → OpenAI | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| Gemini → OpenAI | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| Gemini → OpenAI | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| Gemini → OpenAI | Throughput (req/s) | 68 | concurrency 1; n=410 | 50 ms |
| Gemini → OpenAI | Throughput (req/s) | 68 | concurrency 1; n=410 | 100 ms |
| Gemini → OpenAI | Throughput (req/s) | 68 | concurrency 1; n=410 | Unbounded |
| Cohere → OpenAI Responses | — | — | Not declared: recorded note 1 | — |
| Bedrock → OpenAI Responses | — | — | Not declared: recorded note 1 | — |
| Gemini → OpenAI Responses | Added latency p50 (µs) | 17,577 | Valid · n=338 gateway / 262,004 direct | — |
| Gemini → OpenAI Responses | Added latency p99 (µs) | 18,499 | Valid · n=338 gateway / 262,004 direct | — |
| Gemini → OpenAI Responses | Gateway p99 at concurrency 1 (µs) | 18,527 | Valid · n=338 | — |
| Gemini → OpenAI Responses | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=262,004 | — |
| Gemini → OpenAI Responses | CPU per request (µs) | 18,005.9 | Valid | — |
| Gemini → OpenAI Responses | Requests per CPU-second | 55.54 | Valid | — |
| Gemini → OpenAI Responses | Cost window concurrency | 8 | Valid | — |
| Gemini → OpenAI Responses | Cost window core utilisation | 0.251 | Valid | — |
| Gemini → OpenAI Responses | Cost window successful requests | 341 | Valid | — |
| Gemini → OpenAI Responses | Cost window throughput (req/s) | 56 | Valid | — |
| Gemini → OpenAI Responses | Cost window threads | 13 | Valid | — |
| Gemini → OpenAI Responses | Non-voluntary context switches per request | 0.082 | Valid | — |
| Gemini → OpenAI Responses | Cost window major page faults | 0 | Valid | — |
| Gemini → OpenAI Responses | Added time to first event p50 (µs) | — | Untestable | — |
| Gemini → OpenAI Responses | Added time to first event p99 (µs) | — | Untestable | — |
| Gemini → OpenAI Responses | Added inter-frame gap p50 (µs) | — | Untestable | — |
| Gemini → OpenAI Responses | Added inter-frame gap p99 (µs) | — | Untestable | — |
| Gemini → OpenAI Responses | Streams sustained (frames/s) | — | Untestable | — |
| Gemini → OpenAI Responses | Streams sustained | — | Untestable | — |
| Gemini → OpenAI Responses | Idle RSS (MiB) | 592.4 | Valid | — |
| Gemini → OpenAI Responses | Steady-state RSS (MiB) | 674.6 | Valid | — |
| Gemini → OpenAI Responses | Recovered RSS (MiB) | 674.6 | Valid | — |
| Gemini → OpenAI Responses | Peak RSS (MiB) | 674.6 | Valid | — |
| Gemini → OpenAI Responses | Peak RSS high-water mark (MiB) | 674.6 | Valid | — |
| Gemini → OpenAI Responses | Time to plateau (s) | 63 | Valid | — |
| Gemini → OpenAI Responses | Memory growth (MiB/min) | 0 | Valid | — |
| Gemini → OpenAI Responses | Memory load duration (s) | 306 | Valid | — |
| Gemini → OpenAI Responses | Memory shape under load | — | Not measured | — |
| Gemini → OpenAI Responses | Idle memory shape | — | Not measured | — |
| Gemini → OpenAI Responses | Idle memory static | 1 | Valid | — |
| Gemini → OpenAI Responses | Idle memory growth (MiB/min) | 0 | Valid | — |
| Gemini → OpenAI Responses | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| Gemini → OpenAI Responses | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| Gemini → OpenAI Responses | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| Gemini → OpenAI Responses | Throughput (req/s) | 56 | concurrency 1; n=337 | 50 ms |
| Gemini → OpenAI Responses | Throughput (req/s) | 56 | concurrency 1; n=337 | 100 ms |
| Gemini → OpenAI Responses | Throughput (req/s) | 59 | concurrency 2,048; n=360 | Unbounded |
| Anthropic → OpenAI Responses | Added latency p50 (µs) | 13,287 | Valid · n=449 gateway / 260,854 direct | — |
| Anthropic → OpenAI Responses | Added latency p99 (µs) | 14,010 | Valid · n=449 gateway / 260,854 direct | — |
| Anthropic → OpenAI Responses | Gateway p99 at concurrency 1 (µs) | 14,038 | Valid · n=449 | — |
| Anthropic → OpenAI Responses | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=260,854 | — |
| Anthropic → OpenAI Responses | CPU per request (µs) | 13,738.7 | Valid | — |
| Anthropic → OpenAI Responses | Requests per CPU-second | 72.79 | Valid | — |
| Anthropic → OpenAI Responses | Cost window concurrency | 8 | Valid | — |
| Anthropic → OpenAI Responses | Cost window core utilisation | 0.25 | Valid | — |
| Anthropic → OpenAI Responses | Cost window successful requests | 444 | Valid | — |
| Anthropic → OpenAI Responses | Cost window throughput (req/s) | 73 | Valid | — |
| Anthropic → OpenAI Responses | Cost window threads | 13 | Valid | — |
| Anthropic → OpenAI Responses | Non-voluntary context switches per request | 0.072 | Valid | — |
| Anthropic → OpenAI Responses | Cost window major page faults | 0 | Valid | — |
| Anthropic → OpenAI Responses | Added time to first event p50 (µs) | — | Untestable | — |
| Anthropic → OpenAI Responses | Added time to first event p99 (µs) | — | Untestable | — |
| Anthropic → OpenAI Responses | Added inter-frame gap p50 (µs) | — | Untestable | — |
| Anthropic → OpenAI Responses | Added inter-frame gap p99 (µs) | — | Untestable | — |
| Anthropic → OpenAI Responses | Streams sustained (frames/s) | — | Untestable | — |
| Anthropic → OpenAI Responses | Streams sustained | — | Untestable | — |
| Anthropic → OpenAI Responses | Idle RSS (MiB) | 591.7 | Valid | — |
| Anthropic → OpenAI Responses | Steady-state RSS (MiB) | 661.2 | Valid | — |
| Anthropic → OpenAI Responses | Recovered RSS (MiB) | 661.2 | Valid | — |
| Anthropic → OpenAI Responses | Peak RSS (MiB) | 661.2 | Valid | — |
| Anthropic → OpenAI Responses | Peak RSS high-water mark (MiB) | 661.2 | Valid | — |
| Anthropic → OpenAI Responses | Time to plateau (s) | 63 | Valid | — |
| Anthropic → OpenAI Responses | Memory growth (MiB/min) | 0 | Valid | — |
| Anthropic → OpenAI Responses | Memory load duration (s) | 303 | Valid | — |
| Anthropic → OpenAI Responses | Memory shape under load | — | Not measured | — |
| Anthropic → OpenAI Responses | Idle memory shape | — | Not measured | — |
| Anthropic → OpenAI Responses | Idle memory static | 1 | Valid | — |
| Anthropic → OpenAI Responses | Idle memory growth (MiB/min) | 0 | Valid | — |
| Anthropic → OpenAI Responses | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| Anthropic → OpenAI Responses | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| Anthropic → OpenAI Responses | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| Anthropic → OpenAI Responses | Throughput (req/s) | 74 | concurrency 1; n=450 | 50 ms |
| Anthropic → OpenAI Responses | Throughput (req/s) | 74 | concurrency 1; n=450 | 100 ms |
| Anthropic → OpenAI Responses | Throughput (req/s) | 74 | concurrency 1; n=450 | Unbounded |
| OpenAI → OpenAI Responses | Added latency p50 (µs) | 10,726 | Valid · n=557 gateway / 261,925 direct | — |
| OpenAI → OpenAI Responses | Added latency p99 (µs) | 11,474 | Valid · n=557 gateway / 261,925 direct | — |
| OpenAI → OpenAI Responses | Gateway p99 at concurrency 1 (µs) | 11,502 | Valid · n=557 | — |
| OpenAI → OpenAI Responses | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=261,925 | — |
| OpenAI → OpenAI Responses | CPU per request (µs) | 10,799.3 | Valid | — |
| OpenAI → OpenAI Responses | Requests per CPU-second | 92.6 | Valid | — |
| OpenAI → OpenAI Responses | Cost window concurrency | 8 | Valid | — |
| OpenAI → OpenAI Responses | Cost window core utilisation | 0.25 | Valid | — |
| OpenAI → OpenAI Responses | Cost window successful requests | 563 | Valid | — |
| OpenAI → OpenAI Responses | Cost window throughput (req/s) | 93 | Valid | — |
| OpenAI → OpenAI Responses | Cost window threads | 13 | Valid | — |
| OpenAI → OpenAI Responses | Non-voluntary context switches per request | 0.053 | Valid | — |
| OpenAI → OpenAI Responses | Cost window major page faults | 0 | Valid | — |
| OpenAI → OpenAI Responses | Added time to first event p50 (µs) | — | Untestable | — |
| OpenAI → OpenAI Responses | Added time to first event p99 (µs) | — | Untestable | — |
| OpenAI → OpenAI Responses | Added inter-frame gap p50 (µs) | — | Untestable | — |
| OpenAI → OpenAI Responses | Added inter-frame gap p99 (µs) | — | Untestable | — |
| OpenAI → OpenAI Responses | Streams sustained (frames/s) | — | Untestable | — |
| OpenAI → OpenAI Responses | Streams sustained | — | Untestable | — |
| OpenAI → OpenAI Responses | Idle RSS (MiB) | 591.8 | Valid | — |
| OpenAI → OpenAI Responses | Steady-state RSS (MiB) | 669.4 | Valid | — |
| OpenAI → OpenAI Responses | Recovered RSS (MiB) | 669.4 | Valid | — |
| OpenAI → OpenAI Responses | Peak RSS (MiB) | 669.4 | Valid | — |
| OpenAI → OpenAI Responses | Peak RSS high-water mark (MiB) | 669.4 | Valid | — |
| OpenAI → OpenAI Responses | Time to plateau (s) | 62 | Valid | — |
| OpenAI → OpenAI Responses | Memory growth (MiB/min) | 0 | Valid | — |
| OpenAI → OpenAI Responses | Memory load duration (s) | 305 | Valid | — |
| OpenAI → OpenAI Responses | Memory shape under load | — | Not measured | — |
| OpenAI → OpenAI Responses | Idle memory shape | — | Not measured | — |
| OpenAI → OpenAI Responses | Idle memory static | 1 | Valid | — |
| OpenAI → OpenAI Responses | Idle memory growth (MiB/min) | 0 | Valid | — |
| OpenAI → OpenAI Responses | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| OpenAI → OpenAI Responses | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| OpenAI → OpenAI Responses | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| OpenAI → OpenAI Responses | Throughput (req/s) | 92 | concurrency 1; n=556 | 50 ms |
| OpenAI → OpenAI Responses | Throughput (req/s) | 92 | concurrency 1; n=556 | 100 ms |
| OpenAI → OpenAI Responses | Throughput (req/s) | 93 | concurrency 16; n=559 | Unbounded |
| OpenAI Responses → OpenAI Responses | Added latency p50 (µs) | 11,554 | Valid · n=518 gateway / 265,921 direct | — |
| OpenAI Responses → OpenAI Responses | Added latency p99 (µs) | 12,280 | Valid · n=518 gateway / 265,921 direct | — |
| OpenAI Responses → OpenAI Responses | Gateway p99 at concurrency 1 (µs) | 12,308 | Valid · n=518 | — |
| OpenAI Responses → OpenAI Responses | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=265,921 | — |
| OpenAI Responses → OpenAI Responses | CPU per request (µs) | 11,692.3 | Valid | — |
| OpenAI Responses → OpenAI Responses | Requests per CPU-second | 85.53 | Valid | — |
| OpenAI Responses → OpenAI Responses | Cost window concurrency | 8 | Valid | — |
| OpenAI Responses → OpenAI Responses | Cost window core utilisation | 0.25 | Valid | — |
| OpenAI Responses → OpenAI Responses | Cost window successful requests | 520 | Valid | — |
| OpenAI Responses → OpenAI Responses | Cost window throughput (req/s) | 86 | Valid | — |
| OpenAI Responses → OpenAI Responses | Cost window threads | 13 | Valid | — |
| OpenAI Responses → OpenAI Responses | Non-voluntary context switches per request | 0.063 | Valid | — |
| OpenAI Responses → OpenAI Responses | Cost window major page faults | 0 | Valid | — |
| OpenAI Responses → OpenAI Responses | Added time to first event p50 (µs) | — | Untestable | — |
| OpenAI Responses → OpenAI Responses | Added time to first event p99 (µs) | — | Untestable | — |
| OpenAI Responses → OpenAI Responses | Added inter-frame gap p50 (µs) | — | Untestable | — |
| OpenAI Responses → OpenAI Responses | Added inter-frame gap p99 (µs) | — | Untestable | — |
| OpenAI Responses → OpenAI Responses | Streams sustained (frames/s) | — | Untestable | — |
| OpenAI Responses → OpenAI Responses | Streams sustained | — | Untestable | — |
| OpenAI Responses → OpenAI Responses | Idle RSS (MiB) | 592.5 | Valid | — |
| OpenAI Responses → OpenAI Responses | Steady-state RSS (MiB) | 627.4 | Valid | — |
| OpenAI Responses → OpenAI Responses | Recovered RSS (MiB) | 629.5 | Valid | — |
| OpenAI Responses → OpenAI Responses | Peak RSS (MiB) | 629.5 | Valid | — |
| OpenAI Responses → OpenAI Responses | Peak RSS high-water mark (MiB) | 629.5 | Valid | — |
| OpenAI Responses → OpenAI Responses | Time to plateau (s) | 62 | Valid | — |
| OpenAI Responses → OpenAI Responses | Memory growth (MiB/min) | 3.04 | Valid | — |
| OpenAI Responses → OpenAI Responses | Memory load duration (s) | 300 | Valid | — |
| OpenAI Responses → OpenAI Responses | Memory shape under load | — | Not measured | — |
| OpenAI Responses → OpenAI Responses | Idle memory shape | — | Not measured | — |
| OpenAI Responses → OpenAI Responses | Idle memory static | 1 | Valid | — |
| OpenAI Responses → OpenAI Responses | Idle memory growth (MiB/min) | 0 | Valid | — |
| OpenAI Responses → OpenAI Responses | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| OpenAI Responses → OpenAI Responses | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| OpenAI Responses → OpenAI Responses | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| OpenAI Responses → OpenAI Responses | Throughput (req/s) | 87 | concurrency 2; n=524 | 50 ms |
| OpenAI Responses → OpenAI Responses | Throughput (req/s) | 87 | concurrency 2; n=524 | 100 ms |
| OpenAI Responses → OpenAI Responses | Throughput (req/s) | 88 | concurrency 32; n=529 | Unbounded |
| OpenAI Responses → Gemini | Added latency p50 (µs) | 10,486 | Valid · n=570 gateway / 265,239 direct | — |
| OpenAI Responses → Gemini | Added latency p99 (µs) | 11,201 | Valid · n=570 gateway / 265,239 direct | — |
| OpenAI Responses → Gemini | Gateway p99 at concurrency 1 (µs) | 11,229 | Valid · n=570 | — |
| OpenAI Responses → Gemini | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=265,239 | — |
| OpenAI Responses → Gemini | CPU per request (µs) | 10,628.3 | Valid | — |
| OpenAI Responses → Gemini | Requests per CPU-second | 94.09 | Valid | — |
| OpenAI Responses → Gemini | Cost window concurrency | 8 | Valid | — |
| OpenAI Responses → Gemini | Cost window core utilisation | 0.25 | Valid | — |
| OpenAI Responses → Gemini | Cost window successful requests | 573 | Valid | — |
| OpenAI Responses → Gemini | Cost window throughput (req/s) | 95 | Valid | — |
| OpenAI Responses → Gemini | Cost window threads | 13 | Valid | — |
| OpenAI Responses → Gemini | Non-voluntary context switches per request | 0.042 | Valid | — |
| OpenAI Responses → Gemini | Cost window major page faults | 0 | Valid | — |
| OpenAI Responses → Gemini | Added time to first event p50 (µs) | — | Untestable | — |
| OpenAI Responses → Gemini | Added time to first event p99 (µs) | — | Untestable | — |
| OpenAI Responses → Gemini | Added inter-frame gap p50 (µs) | — | Untestable | — |
| OpenAI Responses → Gemini | Added inter-frame gap p99 (µs) | — | Untestable | — |
| OpenAI Responses → Gemini | Streams sustained (frames/s) | — | Untestable | — |
| OpenAI Responses → Gemini | Streams sustained | — | Untestable | — |
| OpenAI Responses → Gemini | Idle RSS (MiB) | 592.4 | Valid | — |
| OpenAI Responses → Gemini | Steady-state RSS (MiB) | 669.4 | Valid | — |
| OpenAI Responses → Gemini | Recovered RSS (MiB) | 669.4 | Valid | — |
| OpenAI Responses → Gemini | Peak RSS (MiB) | 669.4 | Valid | — |
| OpenAI Responses → Gemini | Peak RSS high-water mark (MiB) | 669.4 | Valid | — |
| OpenAI Responses → Gemini | Time to plateau (s) | 62 | Valid | — |
| OpenAI Responses → Gemini | Memory growth (MiB/min) | 0 | Valid | — |
| OpenAI Responses → Gemini | Memory load duration (s) | 304 | Valid | — |
| OpenAI Responses → Gemini | Memory shape under load | — | Not measured | — |
| OpenAI Responses → Gemini | Idle memory shape | — | Not measured | — |
| OpenAI Responses → Gemini | Idle memory static | 1 | Valid | — |
| OpenAI Responses → Gemini | Idle memory growth (MiB/min) | 0 | Valid | — |
| OpenAI Responses → Gemini | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| OpenAI Responses → Gemini | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| OpenAI Responses → Gemini | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| OpenAI Responses → Gemini | Throughput (req/s) | 96 | concurrency 2; n=578 | 50 ms |
| OpenAI Responses → Gemini | Throughput (req/s) | 96 | concurrency 2; n=578 | 100 ms |
| OpenAI Responses → Gemini | Throughput (req/s) | 96 | concurrency 2; n=578 | Unbounded |
| Gemini → Gemini | Added latency p50 (µs) | 10,073 | Valid · n=592 gateway / 267,671 direct | — |
| Gemini → Gemini | Added latency p99 (µs) | 10,889 | Valid · n=592 gateway / 267,671 direct | — |
| Gemini → Gemini | Gateway p99 at concurrency 1 (µs) | 10,916 | Valid · n=592 | — |
| Gemini → Gemini | Direct-to-mock p99 at concurrency 1 (µs) | 27 | Valid · n=267,671 | — |
| Gemini → Gemini | CPU per request (µs) | 10,236.5 | Valid | — |
| Gemini → Gemini | Requests per CPU-second | 97.69 | Valid | — |
| Gemini → Gemini | Cost window concurrency | 8 | Valid | — |
| Gemini → Gemini | Cost window core utilisation | 0.25 | Valid | — |
| Gemini → Gemini | Cost window successful requests | 592 | Valid | — |
| Gemini → Gemini | Cost window throughput (req/s) | 98 | Valid | — |
| Gemini → Gemini | Cost window threads | 13 | Valid | — |
| Gemini → Gemini | Non-voluntary context switches per request | 0.049 | Valid | — |
| Gemini → Gemini | Cost window major page faults | 0 | Valid | — |
| Gemini → Gemini | Added time to first event p50 (µs) | — | Untestable | — |
| Gemini → Gemini | Added time to first event p99 (µs) | — | Untestable | — |
| Gemini → Gemini | Added inter-frame gap p50 (µs) | — | Untestable | — |
| Gemini → Gemini | Added inter-frame gap p99 (µs) | — | Untestable | — |
| Gemini → Gemini | Streams sustained (frames/s) | — | Untestable | — |
| Gemini → Gemini | Streams sustained | — | Untestable | — |
| Gemini → Gemini | Idle RSS (MiB) | 591.5 | Valid | — |
| Gemini → Gemini | Steady-state RSS (MiB) | 627.7 | Valid | — |
| Gemini → Gemini | Recovered RSS (MiB) | 627.7 | Valid | — |
| Gemini → Gemini | Peak RSS (MiB) | 627.7 | Valid | — |
| Gemini → Gemini | Peak RSS high-water mark (MiB) | 627.7 | Valid | — |
| Gemini → Gemini | Time to plateau (s) | 61 | Valid | — |
| Gemini → Gemini | Memory growth (MiB/min) | 0 | Valid | — |
| Gemini → Gemini | Memory load duration (s) | 302 | Valid | — |
| Gemini → Gemini | Memory shape under load | — | Not measured | — |
| Gemini → Gemini | Idle memory shape | — | Not measured | — |
| Gemini → Gemini | Idle memory static | 1 | Valid | — |
| Gemini → Gemini | Idle memory growth (MiB/min) | 0 | Valid | — |
| Gemini → Gemini | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| Gemini → Gemini | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| Gemini → Gemini | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 10ms, so this gateway carried no measurable throughput under that bound | 10 ms |
| Gemini → Gemini | Throughput (req/s) | 101 | concurrency 2; n=612 | 50 ms |
| Gemini → Gemini | Throughput (req/s) | 101 | concurrency 2; n=612 | 100 ms |
| Gemini → Gemini | Throughput (req/s) | 101 | concurrency 2; n=612 | Unbounded |
| Cohere → Gemini | — | — | Not declared: recorded note 1 | — |
| Bedrock → Gemini | — | — | Not declared: recorded note 1 | — |
| OpenAI → Gemini | Added latency p50 (µs) | 8,225 | Valid · n=725 gateway / 265,737 direct | — |
| OpenAI → Gemini | Added latency p99 (µs) | 8,995 | Valid · n=725 gateway / 265,737 direct | — |
| OpenAI → Gemini | Gateway p99 at concurrency 1 (µs) | 9,023 | Valid · n=725 | — |
| OpenAI → Gemini | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=265,737 | — |
| OpenAI → Gemini | CPU per request (µs) | 8,281 | Valid | — |
| OpenAI → Gemini | Requests per CPU-second | 120.8 | Valid | — |
| OpenAI → Gemini | Cost window concurrency | 8 | Valid | — |
| OpenAI → Gemini | Cost window core utilisation | 0.25 | Valid | — |
| OpenAI → Gemini | Cost window successful requests | 733 | Valid | — |
| OpenAI → Gemini | Cost window throughput (req/s) | 122 | Valid | — |
| OpenAI → Gemini | Cost window threads | 13 | Valid | — |
| OpenAI → Gemini | Non-voluntary context switches per request | 0.029 | Valid | — |
| OpenAI → Gemini | Cost window major page faults | 0 | Valid | — |
| OpenAI → Gemini | Added time to first event p50 (µs) | — | Untestable | — |
| OpenAI → Gemini | Added time to first event p99 (µs) | — | Untestable | — |
| OpenAI → Gemini | Added inter-frame gap p50 (µs) | — | Untestable | — |
| OpenAI → Gemini | Added inter-frame gap p99 (µs) | — | Untestable | — |
| OpenAI → Gemini | Streams sustained (frames/s) | — | Untestable | — |
| OpenAI → Gemini | Streams sustained | — | Untestable | — |
| OpenAI → Gemini | Idle RSS (MiB) | 591.8 | Valid | — |
| OpenAI → Gemini | Steady-state RSS (MiB) | 665.4 | Valid | — |
| OpenAI → Gemini | Recovered RSS (MiB) | 665.4 | Valid | — |
| OpenAI → Gemini | Peak RSS (MiB) | 665.4 | Valid | — |
| OpenAI → Gemini | Peak RSS high-water mark (MiB) | 665.4 | Valid | — |
| OpenAI → Gemini | Time to plateau (s) | 61 | Valid | — |
| OpenAI → Gemini | Memory growth (MiB/min) | 0 | Valid | — |
| OpenAI → Gemini | Memory load duration (s) | 302 | Valid | — |
| OpenAI → Gemini | Memory shape under load | — | Not measured | — |
| OpenAI → Gemini | Idle memory shape | — | Not measured | — |
| OpenAI → Gemini | Idle memory static | 1 | Valid | — |
| OpenAI → Gemini | Idle memory growth (MiB/min) | 0 | Valid | — |
| OpenAI → Gemini | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| OpenAI → Gemini | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| OpenAI → Gemini | Throughput (req/s) | 121 | concurrency 1; n=732 | 10 ms |
| OpenAI → Gemini | Throughput (req/s) | 123 | concurrency 2; n=740 | 50 ms |
| OpenAI → Gemini | Throughput (req/s) | 123 | concurrency 2; n=740 | 100 ms |
| OpenAI → Gemini | Throughput (req/s) | 127 | concurrency 128; n=767 | Unbounded |
| Anthropic → Gemini | Added latency p50 (µs) | 8,964 | Valid · n=666 gateway / 263,751 direct | — |
| Anthropic → Gemini | Added latency p99 (µs) | 9,686 | Valid · n=666 gateway / 263,751 direct | — |
| Anthropic → Gemini | Gateway p99 at concurrency 1 (µs) | 9,714 | Valid · n=666 | — |
| Anthropic → Gemini | Direct-to-mock p99 at concurrency 1 (µs) | 28 | Valid · n=263,751 | — |
| Anthropic → Gemini | CPU per request (µs) | 9,144.1 | Valid | — |
| Anthropic → Gemini | Requests per CPU-second | 109.4 | Valid | — |
| Anthropic → Gemini | Cost window concurrency | 8 | Valid | — |
| Anthropic → Gemini | Cost window core utilisation | 0.251 | Valid | — |
| Anthropic → Gemini | Cost window successful requests | 666 | Valid | — |
| Anthropic → Gemini | Cost window throughput (req/s) | 110 | Valid | — |
| Anthropic → Gemini | Cost window threads | 13 | Valid | — |
| Anthropic → Gemini | Non-voluntary context switches per request | 0.035 | Valid | — |
| Anthropic → Gemini | Cost window major page faults | 0 | Valid | — |
| Anthropic → Gemini | Added time to first event p50 (µs) | — | Untestable | — |
| Anthropic → Gemini | Added time to first event p99 (µs) | — | Untestable | — |
| Anthropic → Gemini | Added inter-frame gap p50 (µs) | — | Untestable | — |
| Anthropic → Gemini | Added inter-frame gap p99 (µs) | — | Untestable | — |
| Anthropic → Gemini | Streams sustained (frames/s) | — | Untestable | — |
| Anthropic → Gemini | Streams sustained | — | Untestable | — |
| Anthropic → Gemini | Idle RSS (MiB) | 593 | Valid | — |
| Anthropic → Gemini | Steady-state RSS (MiB) | 666.6 | Valid | — |
| Anthropic → Gemini | Recovered RSS (MiB) | 666.6 | Valid | — |
| Anthropic → Gemini | Peak RSS (MiB) | 666.6 | Valid | — |
| Anthropic → Gemini | Peak RSS high-water mark (MiB) | 666.6 | Valid | — |
| Anthropic → Gemini | Time to plateau (s) | 61 | Valid | — |
| Anthropic → Gemini | Memory growth (MiB/min) | 0.006 | Valid | — |
| Anthropic → Gemini | Memory load duration (s) | 302 | Valid | — |
| Anthropic → Gemini | Memory shape under load | — | Not measured | — |
| Anthropic → Gemini | Idle memory shape | — | Not measured | — |
| Anthropic → Gemini | Idle memory static | 1 | Valid | — |
| Anthropic → Gemini | Idle memory growth (MiB/min) | 0 | Valid | — |
| Anthropic → Gemini | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 1ms, so this gateway carried no measurable throughput under that bound | 1 ms |
| Anthropic → Gemini | Throughput (req/s) | — | Bound not met; concurrency not recorded: every cleanly-served rung that reported a tail had it at or above 5ms, so this gateway carried no measurable throughput under that bound | 5 ms |
| Anthropic → Gemini | Throughput (req/s) | 111 | concurrency 1; n=667 | 10 ms |
| Anthropic → Gemini | Throughput (req/s) | 113 | concurrency 2; n=682 | 50 ms |
| Anthropic → Gemini | Throughput (req/s) | 113 | concurrency 2; n=682 | 100 ms |
| Anthropic → Gemini | Throughput (req/s) | 113 | concurrency 2; n=682 | Unbounded |
Probe evidence and instrument provenance
Historical capability notes may reference an earlier product version. They are not a new probe or a current support claim. Full RSS samples remain in the immutable raw JSON.
| Ingress → upstream | Cell state | Probe HTTP status | Recorded reason | Recorded detail | Evidence scope | Response excerpt | Tested build |
|---|---|---|---|---|---|---|---|
| Cohere → Cohere | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| OpenAI → Cohere | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| Bedrock → Cohere | Not declared | — | recorded note 1 | recorded note 1 | recorded note 2 | — | ghcr.io/berriai/litellm:v1.108.3 |
| Anthropic → Cohere | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| OpenAI Responses → Cohere | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| Gemini → Cohere | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| Cohere → Anthropic | Not declared | — | recorded note 1 | recorded note 1 | recorded note 2 | — | ghcr.io/berriai/litellm:v1.108.3 |
| Bedrock → Anthropic | Not declared | — | recorded note 1 | recorded note 1 | recorded note 2 | — | ghcr.io/berriai/litellm:v1.108.3 |
| OpenAI Responses → Anthropic | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| Gemini → Anthropic | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| OpenAI → Anthropic | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| Anthropic → Anthropic | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| OpenAI Responses → Bedrock | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| Cohere → Bedrock | Not declared | — | recorded note 1 | recorded note 1 | recorded note 2 | — | ghcr.io/berriai/litellm:v1.108.3 |
| Anthropic → Bedrock | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| Gemini → Bedrock | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| Bedrock → Bedrock | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| OpenAI → Bedrock | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| OpenAI → OpenAI | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| Anthropic → OpenAI | Not served | 200 | failed | probed and answered HTTP 200, but the body is not a usable anthropic response (see body_snippet) | Recorded run evidence | the 2xx body is not a response of the anthropic dialect with assistant text at content[type=text].text: {"id":"resp_bGl0ZWxsbTpjdXN0b21fbGxtX3Byb3ZpZGVyOm9wZW5haTttb2RlbF9pZDo2MTk3ZDQ1OWNlMzc1OTE2NzVlMTIyMTg3ZDAwOTViNzIyNDhmMTRmYTUwNzQzMDZkMzFlMzlhODkzMTM2OWQ4O3Jlc3BvbnNlX2lkOmNoYXRjbXBsLXg=","type":"message","role":"assistant","model":"gpt-4o-mini","stop_sequence":null,"usage":{"input_tokens":10,"output_tokens":2},"content":[],"stop_reason":"end_turn","stop_details":null} | ghcr.io/berriai/litellm:v1.108.3 |
| OpenAI Responses → OpenAI | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| Cohere → OpenAI | Not declared | — | recorded note 1 | recorded note 1 | recorded note 2 | — | ghcr.io/berriai/litellm:v1.108.3 |
| Bedrock → OpenAI | Not declared | — | recorded note 1 | recorded note 1 | recorded note 2 | — | ghcr.io/berriai/litellm:v1.108.3 |
| Gemini → OpenAI | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| Cohere → OpenAI Responses | Not declared | — | recorded note 1 | recorded note 1 | recorded note 2 | — | ghcr.io/berriai/litellm:v1.108.3 |
| Bedrock → OpenAI Responses | Not declared | — | recorded note 1 | recorded note 1 | recorded note 2 | — | ghcr.io/berriai/litellm:v1.108.3 |
| Gemini → OpenAI Responses | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| Anthropic → OpenAI Responses | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| OpenAI → OpenAI Responses | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| OpenAI Responses → OpenAI Responses | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| OpenAI Responses → Gemini | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| Gemini → Gemini | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| Cohere → Gemini | Not declared | — | recorded note 1 | recorded note 1 | recorded note 2 | — | ghcr.io/berriai/litellm:v1.108.3 |
| Bedrock → Gemini | Not declared | — | recorded note 1 | recorded note 1 | recorded note 2 | — | ghcr.io/berriai/litellm:v1.108.3 |
| OpenAI → Gemini | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
| Anthropic → Gemini | Served | — | — | — | Recorded run evidence | — | ghcr.io/berriai/litellm:v1.108.3 |
- Harness commit
d240badd966933fa19eec0606b2d1f3945f18ad2- Mock SHA-256
99d01f709d8d9c1f49055aa265d5eb7475264d6b513911fe317aee74a71190f8- Load generator SHA-256
not recorded- net.core.somaxconn
- 65,535
- net.ipv4.tcp_max_syn_backlog
- 65,535
Stream sweep diagnostics
Per streaming-sweep concurrency, as recorded by the engine: lanes by final HTTP status, and lanes that received no response.
| Ingress → upstream | Concurrency | Lanes by HTTP status | Lanes with no response |
|---|---|---|---|
| OpenAI → Anthropic | 1 | HTTP 200: 1 | 0 |
| OpenAI → Anthropic | 2 | HTTP 200: 2 | 0 |
| OpenAI → Anthropic | 4 | HTTP 200: 4 | 0 |
| OpenAI → Anthropic | 8 | HTTP 200: 8 | 0 |
| OpenAI → Anthropic | 16 | HTTP 200: 16 | 0 |
| OpenAI → Anthropic | 32 | HTTP 200: 32 | 0 |
| OpenAI → Anthropic | 64 | HTTP 200: 64 | 0 |
| OpenAI → Anthropic | 128 | HTTP 200: 128 | 0 |
| OpenAI → Anthropic | 96 | HTTP 200: 96 | 0 |
| OpenAI → Anthropic | 112 | HTTP 200: 112 | 0 |
| OpenAI → Anthropic | 120 | HTTP 200: 120 | 0 |
| OpenAI → Anthropic | 116 | HTTP 200: 116 | 0 |
| OpenAI → Anthropic | 118 | HTTP 200: 118 | 0 |
| OpenAI → Anthropic | 117 | HTTP 200: 117 | 0 |
| OpenAI → Anthropic | 117 | HTTP 200: 117 | 0 |
| OpenAI → Anthropic | 117 | HTTP 200: 117 | 0 |
| Anthropic → Anthropic | 1 | HTTP 200: 1 | 0 |
| Anthropic → Anthropic | 2 | HTTP 200: 2 | 0 |
| Anthropic → Anthropic | 4 | HTTP 200: 4 | 0 |
| Anthropic → Anthropic | 8 | HTTP 200: 8 | 0 |
| Anthropic → Anthropic | 16 | HTTP 200: 16 | 0 |
| Anthropic → Anthropic | 32 | HTTP 200: 32 | 0 |
| Anthropic → Anthropic | 64 | HTTP 200: 64 | 0 |
| Anthropic → Anthropic | 128 | HTTP 200: 128 | 0 |
| Anthropic → Anthropic | 256 | HTTP 200: 256 | 0 |
| Anthropic → Anthropic | 192 | HTTP 200: 192 | 0 |
| Anthropic → Anthropic | 224 | HTTP 200: 224 | 0 |
| Anthropic → Anthropic | 240 | HTTP 200: 240 | 0 |
| Anthropic → Anthropic | 248 | HTTP 200: 248 | 0 |
| Anthropic → Anthropic | 252 | HTTP 200: 252 | 0 |
| Anthropic → Anthropic | 254 | HTTP 200: 254 | 0 |
| Anthropic → Anthropic | 255 | HTTP 200: 255 | 0 |
| Anthropic → Anthropic | 255 | HTTP 200: 255 | 0 |
| Anthropic → Anthropic | 255 | HTTP 200: 255 | 0 |
| OpenAI → OpenAI | 1 | HTTP 200: 1 | 0 |
| OpenAI → OpenAI | 2 | HTTP 200: 2 | 0 |
| OpenAI → OpenAI | 4 | HTTP 200: 4 | 0 |
| OpenAI → OpenAI | 8 | HTTP 200: 8 | 0 |
| OpenAI → OpenAI | 16 | HTTP 200: 16 | 0 |
| OpenAI → OpenAI | 32 | HTTP 200: 32 | 0 |
| OpenAI → OpenAI | 64 | HTTP 200: 64 | 0 |
| OpenAI → OpenAI | 48 | HTTP 200: 48 | 0 |
| OpenAI → OpenAI | 56 | HTTP 200: 56 | 0 |
| OpenAI → OpenAI | 60 | HTTP 200: 60 | 0 |
| OpenAI → OpenAI | 62 | HTTP 200: 62 | 0 |
| OpenAI → OpenAI | 63 | HTTP 200: 63 | 0 |
| OpenAI → OpenAI | 62 | HTTP 200: 62 | 0 |
| OpenAI → OpenAI | 62 | HTTP 200: 62 | 0 |
Recorded notes
Text recorded with the cells above. Each note is shown once; rows refer to it by number.
- Pinned v1.108.3 source b7d579fab2818968322ef96e5b8e98dc8178fafc: Chat, Responses, Messages and Gemini ingress translate to all six configured upstream wires. The per-model upstream setting selects Chat versus Responses without changing workload parameters. The default Messages-to-Chat bridge reaches the correct wire but returns empty content; it is configured as shipped, without a global bridge override, so its failure is recorded. Native Cohere and Bedrock ingress use the shipped /cohere and /bedrock provider-prefixed passthrough routes in litellm/proxy/pass_through_endpoints/llm_passthrough_endpoints.py; passthrough is same-dialect only. Configured pairs that fail are recorded as failures. Evidence: results/diagnostics/refresh-configuration/litellm-python-final-config-correctness-preflight.json
- Historical capability research, not a fresh probe of the tested build. Referenced versions may differ from gateway_build; do not infer current capability.
Evidence for the current measurements
- run-litellm-python-10bb57507b26d310b59f12956034785b615cd5338cdaf5cee0e587c350c53acf · complete raw record including RSS series
Configuration recorded by the run
model_list:
- model_name: gpt-4o-mini
litellm_params:
model: openai/gpt-4o-mini
use_chat_completions_api: true
api_base: http://127.0.0.1:8000/v1
api_key: dummy
- model_name: gpt-4o-mini-responses
litellm_params:
model: openai/responses/gpt-4o-mini
api_base: http://127.0.0.1:8000/v1
api_key: dummy
- model_name: gpt-4o-mini-anthropic
litellm_params:
model: anthropic/claude-3-5-sonnet-20241022
api_base: http://127.0.0.1:8000
api_key: dummy
- model_name: gpt-4o-mini-gemini
litellm_params:
model: gemini/gemini-1.5-flash
api_base: http://127.0.0.1:8000
api_key: dummy
- model_name: command-r
litellm_params:
model: cohere_chat/command-r
api_base: http://127.0.0.1:8000/v2/chat
api_key: dummy
- model_name: gpt-4o-mini-bedrock
litellm_params:
model: bedrock/converse/anthropic.claude-3-sonnet-20240229-v1:0
aws_bedrock_runtime_endpoint: http://127.0.0.1:8000
aws_access_key_id: AKIAMOCKACCESSKEY
aws_secret_access_key: mock-secret-access-key
aws_region_name: us-east-1
# Public test token for the local mock rig, never a real provider credential.
general_settings:
master_key: sk-otb-local-mock-benchmark
Compare a shared protocol cell
- LiteLLM · Python vs agentgateway
- LiteLLM · Python vs AISIX (api7)
- LiteLLM · Python vs APISIX
- LiteLLM · Python vs Bifrost
- LiteLLM · Python vs Busbar
- LiteLLM · Python vs GoModel
- LiteLLM · Python vs Helicone
- LiteLLM · Python vs Higress
- LiteLLM · Python vs Kong
- LiteLLM · Python vs LiteLLM · Rust
- LiteLLM · Python vs One-API
- LiteLLM · Python vs Plano
- LiteLLM · Python vs Portkey
- LiteLLM · Python vs TensorZero