# Portkey 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](https://onthebench.ai/gateways/method) .

Node · recorded build: portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3. Measured: 2026-11-10T17:39:55Z.

[JSON](https://onthebench.ai/api/v1/gateways/portkey)  ·  [Markdown](https://onthebench.ai/gateways/portkey.md)  ·  [CSV](https://onthebench.ai/gateways/data.csv?gateway=portkey&compact=1)  ·  [Project source](https://github.com/Portkey-AI/gateway)  ·  [Interactive dashboard](https://onthebench.ai/gateways)

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) |
| --- | --- | --- | --- | --- |
| bedrock → openai | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| gemini → openai | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| openai → openai | added_latency_p50_us | 1196 | valid; gateway_c1_samples=4304; direct_c1_samples=246357 | — |
| openai → openai | added_latency_p99_us | 3902 | valid; gateway_c1_samples=4304; direct_c1_samples=246357 | — |
| openai → openai | gateway_c1_p99_us | 3938 | valid; gateway_c1_samples=4304 | — |
| openai → openai | direct_c1_p99_us | 36 | valid; direct_c1_samples=246357 | — |
| openai → openai | cpu_us_per_request | 1528.3748925193465 | valid | — |
| openai → openai | rps_per_cpu_second | 654.2897327707454 | valid | — |
| openai → openai | cost_window_conc | 8 | valid | — |
| openai → openai | cost_core_utilisation | 0.2950181957762656 | valid | — |
| openai → openai | cost_window_ok | 4652 | valid | — |
| openai → openai | cost_window_rps | 775 | valid | — |
| openai → openai | cost_threads | 22 | valid | — |
| openai → openai | cost_nonvol_ctxt_per_request | 0.007308684436801376 | valid | — |
| openai → openai | cost_majflt | 0 | valid | — |
| openai → openai | added_ttft_p50_us | 28114 | valid; ttft_gw_samples=100; ttft_direct_samples=100 | — |
| openai → openai | added_ttft_p99_us | 29752 | valid; ttft_gw_samples=100; ttft_direct_samples=100 | — |
| openai → openai | added_gap_p50_us | 0 | valid; gap_gw_samples=63; gap_direct_samples=63 | — |
| openai → openai | added_gap_p99_us | 440 | valid; gap_gw_samples=63; gap_direct_samples=63 | — |
| openai → openai | streams_sustained_fps | 4996.208831040471 | valid | — |
| openai → openai | streams_sustained | 1519 | valid | — |
| openai → openai | idle_rss_mib | 125.0703125 | valid | — |
| openai → openai | steady_state_rss_mib | 245.171875 | valid | — |
| openai → openai | recovered_rss_mib | 142.08203125 | valid | — |
| openai → openai | peak_rss_mib | 250.01953125 | valid | — |
| openai → openai | peak_rss_hwm_mib | 263.984375 | valid | — |
| openai → openai | time_to_plateau_s | not_measured | not_measured | — |
| openai → openai | growth_rate_mib_per_min | 0.20228106254272332 | valid | — |
| openai → openai | load_s | 301 | valid | — |
| openai → openai | shape | 0 | valid | — |
| openai → openai | idle_shape | -1 | valid | — |
| openai → openai | idle_static | 0 | valid | — |
| openai → openai | idle_growth_rate_mib_per_min | -13.958014762556811 | valid | — |
| openai → openai | throughput_rps | bound_not_met req/s | 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 |
| openai → openai | throughput_rps | 831 req/s | concurrency 1; samples 4992 | 5 |
| openai → openai | throughput_rps | 835 req/s | concurrency 2; samples 5012 | 10 |
| openai → openai | throughput_rps | 836 req/s | concurrency 8; samples 5018 | 50 |
| openai → openai | throughput_rps | 836 req/s | concurrency 8; samples 5018 | 100 |
| openai → openai | throughput_rps | 836 req/s | concurrency 8; samples 5018 | unbounded |
| openai-responses → openai | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| anthropic → openai | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| cohere → openai | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| gemini → cohere | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| cohere → cohere | added_latency_p50_us | 1128 | valid; gateway_c1_samples=4519; direct_c1_samples=248527 | — |
| cohere → cohere | added_latency_p99_us | 3714 | valid; gateway_c1_samples=4519; direct_c1_samples=248527 | — |
| cohere → cohere | gateway_c1_p99_us | 3750 | valid; gateway_c1_samples=4519 | — |
| cohere → cohere | direct_c1_p99_us | 36 | valid; direct_c1_samples=248527 | — |
| cohere → cohere | cpu_us_per_request | 1331.2265455263678 | valid | — |
| cohere → cohere | rps_per_cpu_second | 751.186943620178 | valid | — |
| cohere → cohere | cost_window_conc | 8 | valid | — |
| cohere → cohere | cost_core_utilisation | 0.27976268545278316 | valid | — |
| cohere → cohere | cost_window_ok | 5063 | valid | — |
| cohere → cohere | cost_window_rps | 843 | valid | — |
| cohere → cohere | cost_threads | 22 | valid | — |
| cohere → cohere | cost_nonvol_ctxt_per_request | 0.004937783922575548 | valid | — |
| cohere → cohere | cost_majflt | 0 | valid | — |
| cohere → cohere | added_ttft_p50_us | untestable | untestable | — |
| cohere → cohere | added_ttft_p99_us | untestable | untestable | — |
| cohere → cohere | added_gap_p50_us | untestable | untestable | — |
| cohere → cohere | added_gap_p99_us | untestable | untestable | — |
| cohere → cohere | streams_sustained_fps | untestable | untestable | — |
| cohere → cohere | streams_sustained | untestable | untestable | — |
| cohere → cohere | idle_rss_mib | 124.9921875 | valid | — |
| cohere → cohere | steady_state_rss_mib | 245.9453125 | valid | — |
| cohere → cohere | recovered_rss_mib | 141.52734375 | valid | — |
| cohere → cohere | peak_rss_mib | 249.91796875 | valid | — |
| cohere → cohere | peak_rss_hwm_mib | 264.00390625 | valid | — |
| cohere → cohere | time_to_plateau_s | not_measured | not_measured | — |
| cohere → cohere | growth_rate_mib_per_min | 0.8062392422532753 | valid | — |
| cohere → cohere | load_s | 301 | valid | — |
| cohere → cohere | shape | 0 | valid | — |
| cohere → cohere | idle_shape | -1 | valid | — |
| cohere → cohere | idle_static | 0 | valid | — |
| cohere → cohere | idle_growth_rate_mib_per_min | -14.173574778061678 | valid | — |
| cohere → cohere | throughput_rps | bound_not_met req/s | 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 |
| cohere → cohere | throughput_rps | 860 req/s | concurrency 2; samples 5161 | 5 |
| cohere → cohere | throughput_rps | 860 req/s | concurrency 2; samples 5161 | 10 |
| cohere → cohere | throughput_rps | 860 req/s | concurrency 2; samples 5161 | 50 |
| cohere → cohere | throughput_rps | 860 req/s | concurrency 2; samples 5161 | 100 |
| cohere → cohere | throughput_rps | 860 req/s | concurrency 2; samples 5161 | unbounded |
| bedrock → cohere | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| openai → cohere | added_latency_p50_us | 1204 | valid; gateway_c1_samples=4255; direct_c1_samples=267416 | — |
| openai → cohere | added_latency_p99_us | 3792 | valid; gateway_c1_samples=4255; direct_c1_samples=267416 | — |
| openai → cohere | gateway_c1_p99_us | 3819 | valid; gateway_c1_samples=4255 | — |
| openai → cohere | direct_c1_p99_us | 27 | valid; direct_c1_samples=267416 | — |
| openai → cohere | cpu_us_per_request | 1383.828045035824 | valid | — |
| openai → cohere | rps_per_cpu_second | 722.6331360946746 | valid | — |
| openai → cohere | cost_window_conc | 8 | valid | — |
| openai → cohere | cost_core_utilisation | 0.2804933340659576 | valid | — |
| openai → cohere | cost_window_ok | 4885 | valid | — |
| openai → cohere | cost_window_rps | 813 | valid | — |
| openai → cohere | cost_threads | 22 | valid | — |
| openai → cohere | cost_nonvol_ctxt_per_request | 0.006345957011258956 | valid | — |
| openai → cohere | cost_majflt | 0 | valid | — |
| openai → cohere | added_ttft_p50_us | untestable | untestable | — |
| openai → cohere | added_ttft_p99_us | untestable | untestable | — |
| openai → cohere | added_gap_p50_us | untestable | untestable | — |
| openai → cohere | added_gap_p99_us | untestable | untestable | — |
| openai → cohere | streams_sustained_fps | untestable | untestable | — |
| openai → cohere | streams_sustained | untestable | untestable | — |
| openai → cohere | idle_rss_mib | 124.9375 | valid | — |
| openai → cohere | steady_state_rss_mib | 244.7734375 | valid | — |
| openai → cohere | recovered_rss_mib | 141.5234375 | valid | — |
| openai → cohere | peak_rss_mib | 250.15625 | valid | — |
| openai → cohere | peak_rss_hwm_mib | 264.8359375 | valid | — |
| openai → cohere | time_to_plateau_s | not_measured | not_measured | — |
| openai → cohere | growth_rate_mib_per_min | 0.6758821056788261 | valid | — |
| openai → cohere | load_s | 301 | valid | — |
| openai → cohere | shape | 0 | valid | — |
| openai → cohere | idle_shape | -1 | valid | — |
| openai → cohere | idle_static | 0 | valid | — |
| openai → cohere | idle_growth_rate_mib_per_min | -14.251585141872061 | valid | — |
| openai → cohere | throughput_rps | bound_not_met req/s | 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 |
| openai → cohere | throughput_rps | 821 req/s | concurrency 2; samples 4929 | 5 |
| openai → cohere | throughput_rps | 828 req/s | concurrency 4; samples 4975 | 10 |
| openai → cohere | throughput_rps | 828 req/s | concurrency 4; samples 4975 | 50 |
| openai → cohere | throughput_rps | 828 req/s | concurrency 4; samples 4975 | 100 |
| openai → cohere | throughput_rps | 828 req/s | concurrency 4; samples 4975 | unbounded |
| openai-responses → cohere | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| anthropic → cohere | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| gemini → anthropic | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| cohere → anthropic | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| bedrock → anthropic | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| anthropic → anthropic | added_latency_p50_us | 1196 | valid; gateway_c1_samples=4332; direct_c1_samples=243420 | — |
| anthropic → anthropic | added_latency_p99_us | 3745 | valid; gateway_c1_samples=4332; direct_c1_samples=243420 | — |
| anthropic → anthropic | gateway_c1_p99_us | 3781 | valid; gateway_c1_samples=4332 | — |
| anthropic → anthropic | direct_c1_p99_us | 36 | valid; direct_c1_samples=243420 | — |
| anthropic → anthropic | cpu_us_per_request | 1549.1750589243625 | valid | — |
| anthropic → anthropic | rps_per_cpu_second | 645.5048409405256 | valid | — |
| anthropic → anthropic | cost_window_conc | 8 | valid | — |
| anthropic → anthropic | cost_core_utilisation | 0.3000290843961638 | valid | — |
| anthropic → anthropic | cost_window_ok | 4667 | valid | — |
| anthropic → anthropic | cost_window_rps | 777 | valid | — |
| anthropic → anthropic | cost_threads | 22 | valid | — |
| anthropic → anthropic | cost_nonvol_ctxt_per_request | 0.006642382686950932 | valid | — |
| anthropic → anthropic | cost_majflt | 0 | valid | — |
| anthropic → anthropic | added_ttft_p50_us | 28106 | valid; ttft_gw_samples=100; ttft_direct_samples=100 | — |
| anthropic → anthropic | added_ttft_p99_us | 29379 | valid; ttft_gw_samples=100; ttft_direct_samples=100 | — |
| anthropic → anthropic | added_gap_p50_us | 7 | valid; gap_gw_samples=63; gap_direct_samples=63 | — |
| anthropic → anthropic | added_gap_p99_us | 400 | valid; gap_gw_samples=63; gap_direct_samples=63 | — |
| anthropic → anthropic | streams_sustained_fps | 4968.797812465507 | valid | — |
| anthropic → anthropic | streams_sustained | 1087 | valid | — |
| anthropic → anthropic | idle_rss_mib | 125.01171875 | valid | — |
| anthropic → anthropic | steady_state_rss_mib | 245.30859375 | valid | — |
| anthropic → anthropic | recovered_rss_mib | 141.72265625 | valid | — |
| anthropic → anthropic | peak_rss_mib | 253.70703125 | valid | — |
| anthropic → anthropic | peak_rss_hwm_mib | 268.375 | valid | — |
| anthropic → anthropic | time_to_plateau_s | not_measured | not_measured | — |
| anthropic → anthropic | growth_rate_mib_per_min | -1.4567790319639966 | valid | — |
| anthropic → anthropic | load_s | 301 | valid | — |
| anthropic → anthropic | shape | 0 | valid | — |
| anthropic → anthropic | idle_shape | -1 | valid | — |
| anthropic → anthropic | idle_static | 0 | valid | — |
| anthropic → anthropic | idle_growth_rate_mib_per_min | -14.091683426869364 | valid | — |
| anthropic → anthropic | throughput_rps | bound_not_met req/s | 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 |
| anthropic → anthropic | throughput_rps | 823 req/s | concurrency 2; samples 4941 | 5 |
| anthropic → anthropic | throughput_rps | 823 req/s | concurrency 2; samples 4941 | 10 |
| anthropic → anthropic | throughput_rps | 823 req/s | concurrency 2; samples 4941 | 50 |
| anthropic → anthropic | throughput_rps | 823 req/s | concurrency 2; samples 4941 | 100 |
| anthropic → anthropic | throughput_rps | 823 req/s | concurrency 2; samples 4941 | unbounded |
| openai → anthropic | added_latency_p50_us | 1238 | valid; gateway_c1_samples=4169; direct_c1_samples=245970 | — |
| openai → anthropic | added_latency_p99_us | 3814 | valid; gateway_c1_samples=4169; direct_c1_samples=245970 | — |
| openai → anthropic | gateway_c1_p99_us | 3850 | valid; gateway_c1_samples=4169 | — |
| openai → anthropic | direct_c1_p99_us | 36 | valid; direct_c1_samples=245970 | — |
| openai → anthropic | cpu_us_per_request | 1583.7204158372042 | valid | — |
| openai → anthropic | rps_per_cpu_second | 631.4245810055866 | valid | — |
| openai → anthropic | cost_window_conc | 8 | valid | — |
| openai → anthropic | cost_core_utilisation | 0.29703957089791033 | valid | — |
| openai → anthropic | cost_window_ok | 4521 | valid | — |
| openai → anthropic | cost_window_rps | 753 | valid | — |
| openai → anthropic | cost_threads | 22 | valid | — |
| openai → anthropic | cost_nonvol_ctxt_per_request | 0.007078080070780801 | valid | — |
| openai → anthropic | cost_majflt | 0 | valid | — |
| openai → anthropic | added_ttft_p50_us | 28321 | valid; ttft_gw_samples=100; ttft_direct_samples=100 | — |
| openai → anthropic | added_ttft_p99_us | 29941 | valid; ttft_gw_samples=100; ttft_direct_samples=100 | — |
| openai → anthropic | added_gap_p50_us | 2 | valid; gap_gw_samples=63; gap_direct_samples=63 | — |
| openai → anthropic | added_gap_p99_us | 363 | valid; gap_gw_samples=63; gap_direct_samples=63 | — |
| openai → anthropic | streams_sustained_fps | 4586.904652284963 | valid | — |
| openai → anthropic | streams_sustained | 1055 | valid | — |
| openai → anthropic | idle_rss_mib | 125.16796875 | valid | — |
| openai → anthropic | steady_state_rss_mib | 246.1328125 | valid | — |
| openai → anthropic | recovered_rss_mib | 141.94921875 | valid | — |
| openai → anthropic | peak_rss_mib | 250.5703125 | valid | — |
| openai → anthropic | peak_rss_hwm_mib | 265.05078125 | valid | — |
| openai → anthropic | time_to_plateau_s | not_measured | not_measured | — |
| openai → anthropic | growth_rate_mib_per_min | 1.690937807428462 | valid | — |
| openai → anthropic | load_s | 301 | valid | — |
| openai → anthropic | shape | 0 | valid | — |
| openai → anthropic | idle_shape | -1 | valid | — |
| openai → anthropic | idle_static | 0 | valid | — |
| openai → anthropic | idle_growth_rate_mib_per_min | -13.914391835390527 | valid | — |
| openai → anthropic | throughput_rps | bound_not_met req/s | 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 |
| openai → anthropic | throughput_rps | 824 req/s | concurrency 2; samples 4948 | 5 |
| openai → anthropic | throughput_rps | 824 req/s | concurrency 2; samples 4948 | 10 |
| openai → anthropic | throughput_rps | 824 req/s | concurrency 2; samples 4948 | 50 |
| openai → anthropic | throughput_rps | 824 req/s | concurrency 2; samples 4948 | 100 |
| openai → anthropic | throughput_rps | 824 req/s | concurrency 2; samples 4948 | unbounded |
| openai-responses → anthropic | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| gemini → bedrock | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| bedrock → bedrock | added_latency_p50_us | 1316 | valid; gateway_c1_samples=3896; direct_c1_samples=249657 | — |
| bedrock → bedrock | added_latency_p99_us | 4150 | valid; gateway_c1_samples=3896; direct_c1_samples=249657 | — |
| bedrock → bedrock | gateway_c1_p99_us | 4186 | valid; gateway_c1_samples=3896 | — |
| bedrock → bedrock | direct_c1_p99_us | 36 | valid; direct_c1_samples=249657 | — |
| bedrock → bedrock | cpu_us_per_request | 1534.8314606741574 | valid | — |
| bedrock → bedrock | rps_per_cpu_second | 651.5373352855052 | valid | — |
| bedrock → bedrock | cost_window_conc | 8 | valid | — |
| bedrock → bedrock | cost_core_utilisation | 0.28341991320088467 | valid | — |
| bedrock → bedrock | cost_window_ok | 4450 | valid | — |
| bedrock → bedrock | cost_window_rps | 741 | valid | — |
| bedrock → bedrock | cost_threads | 22 | valid | — |
| bedrock → bedrock | cost_nonvol_ctxt_per_request | 0.00651685393258427 | valid | — |
| bedrock → bedrock | cost_majflt | 0 | valid | — |
| bedrock → bedrock | added_ttft_p50_us | untestable | untestable | — |
| bedrock → bedrock | added_ttft_p99_us | untestable | untestable | — |
| bedrock → bedrock | added_gap_p50_us | untestable | untestable | — |
| bedrock → bedrock | added_gap_p99_us | untestable | untestable | — |
| bedrock → bedrock | streams_sustained_fps | untestable | untestable | — |
| bedrock → bedrock | streams_sustained | untestable | untestable | — |
| bedrock → bedrock | idle_rss_mib | 125.0234375 | valid | — |
| bedrock → bedrock | steady_state_rss_mib | 248.47265625 | valid | — |
| bedrock → bedrock | recovered_rss_mib | 142.32421875 | valid | — |
| bedrock → bedrock | peak_rss_mib | 253.05859375 | valid | — |
| bedrock → bedrock | peak_rss_hwm_mib | 268.09765625 | valid | — |
| bedrock → bedrock | time_to_plateau_s | not_measured | not_measured | — |
| bedrock → bedrock | growth_rate_mib_per_min | 0.32678350093744585 | valid | — |
| bedrock → bedrock | load_s | 301 | valid | — |
| bedrock → bedrock | shape | 0 | valid | — |
| bedrock → bedrock | idle_shape | -1 | valid | — |
| bedrock → bedrock | idle_static | 0 | valid | — |
| bedrock → bedrock | idle_growth_rate_mib_per_min | -14.193759845506435 | valid | — |
| bedrock → bedrock | throughput_rps | bound_not_met req/s | 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 |
| bedrock → bedrock | throughput_rps | 757 req/s | concurrency 2; samples 4547 | 5 |
| bedrock → bedrock | throughput_rps | 757 req/s | concurrency 2; samples 4547 | 10 |
| bedrock → bedrock | throughput_rps | 757 req/s | concurrency 2; samples 4547 | 50 |
| bedrock → bedrock | throughput_rps | 757 req/s | concurrency 2; samples 4547 | 100 |
| bedrock → bedrock | throughput_rps | 757 req/s | concurrency 2; samples 4547 | unbounded |
| openai-responses → bedrock | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| anthropic → bedrock | added_latency_p50_us | 1438 | valid; gateway_c1_samples=3579; direct_c1_samples=262193 | — |
| anthropic → bedrock | added_latency_p99_us | 4467 | valid; gateway_c1_samples=3579; direct_c1_samples=262193 | — |
| anthropic → bedrock | gateway_c1_p99_us | 4494 | valid; gateway_c1_samples=3579 | — |
| anthropic → bedrock | direct_c1_p99_us | 27 | valid; direct_c1_samples=262193 | — |
| anthropic → bedrock | cpu_us_per_request | 1632.6039913440732 | valid | — |
| anthropic → bedrock | rps_per_cpu_second | 612.5184094256259 | valid | — |
| anthropic → bedrock | cost_window_conc | 8 | valid | — |
| anthropic → bedrock | cost_core_utilisation | 0.2817735890792074 | valid | — |
| anthropic → bedrock | cost_window_ok | 4159 | valid | — |
| anthropic → bedrock | cost_window_rps | 693 | valid | — |
| anthropic → bedrock | cost_threads | 22 | valid | — |
| anthropic → bedrock | cost_nonvol_ctxt_per_request | 0.006732387593171436 | valid | — |
| anthropic → bedrock | cost_majflt | 0 | valid | — |
| anthropic → bedrock | added_ttft_p50_us | untestable | untestable | — |
| anthropic → bedrock | added_ttft_p99_us | untestable | untestable | — |
| anthropic → bedrock | added_gap_p50_us | untestable | untestable | — |
| anthropic → bedrock | added_gap_p99_us | untestable | untestable | — |
| anthropic → bedrock | streams_sustained_fps | untestable | untestable | — |
| anthropic → bedrock | streams_sustained | untestable | untestable | — |
| anthropic → bedrock | idle_rss_mib | 125.078125 | valid | — |
| anthropic → bedrock | steady_state_rss_mib | 250.26953125 | valid | — |
| anthropic → bedrock | recovered_rss_mib | 142.53125 | valid | — |
| anthropic → bedrock | peak_rss_mib | 254.4296875 | valid | — |
| anthropic → bedrock | peak_rss_hwm_mib | 268.28125 | valid | — |
| anthropic → bedrock | time_to_plateau_s | not_measured | not_measured | — |
| anthropic → bedrock | growth_rate_mib_per_min | -2.6607292570829726 | valid | — |
| anthropic → bedrock | load_s | 301 | valid | — |
| anthropic → bedrock | shape | 0 | valid | — |
| anthropic → bedrock | idle_shape | -1 | valid | — |
| anthropic → bedrock | idle_static | 0 | valid | — |
| anthropic → bedrock | idle_growth_rate_mib_per_min | -13.92566222223078 | valid | — |
| anthropic → bedrock | throughput_rps | bound_not_met req/s | 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 |
| anthropic → bedrock | throughput_rps | 703 req/s | concurrency 1; samples 4223 | 5 |
| anthropic → bedrock | throughput_rps | 706 req/s | concurrency 2; samples 4237 | 10 |
| anthropic → bedrock | throughput_rps | 706 req/s | concurrency 2; samples 4237 | 50 |
| anthropic → bedrock | throughput_rps | 706 req/s | concurrency 2; samples 4237 | 100 |
| anthropic → bedrock | throughput_rps | 706 req/s | concurrency 2; samples 4237 | unbounded |
| cohere → bedrock | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| openai → bedrock | added_latency_p50_us | 1430 | valid; gateway_c1_samples=3563; direct_c1_samples=269629 | — |
| openai → bedrock | added_latency_p99_us | 4577 | valid; gateway_c1_samples=3563; direct_c1_samples=269629 | — |
| openai → bedrock | gateway_c1_p99_us | 4604 | valid; gateway_c1_samples=3563 | — |
| openai → bedrock | direct_c1_p99_us | 27 | valid; direct_c1_samples=269629 | — |
| openai → bedrock | cpu_us_per_request | 1664.2264519277696 | valid | — |
| openai → bedrock | rps_per_cpu_second | 600.8797653958944 | valid | — |
| openai → bedrock | cost_window_conc | 8 | valid | — |
| openai → bedrock | cost_core_utilisation | 0.28285540553056554 | valid | — |
| openai → bedrock | cost_window_ok | 4098 | valid | — |
| openai → bedrock | cost_window_rps | 682 | valid | — |
| openai → bedrock | cost_threads | 22 | valid | — |
| openai → bedrock | cost_nonvol_ctxt_per_request | 0.008296730112249878 | valid | — |
| openai → bedrock | cost_majflt | 0 | valid | — |
| openai → bedrock | added_ttft_p50_us | untestable | untestable | — |
| openai → bedrock | added_ttft_p99_us | untestable | untestable | — |
| openai → bedrock | added_gap_p50_us | untestable | untestable | — |
| openai → bedrock | added_gap_p99_us | untestable | untestable | — |
| openai → bedrock | streams_sustained_fps | untestable | untestable | — |
| openai → bedrock | streams_sustained | untestable | untestable | — |
| openai → bedrock | idle_rss_mib | 125.18359375 | valid | — |
| openai → bedrock | steady_state_rss_mib | 248.43359375 | valid | — |
| openai → bedrock | recovered_rss_mib | 142.8125 | valid | — |
| openai → bedrock | peak_rss_mib | 253.90625 | valid | — |
| openai → bedrock | peak_rss_hwm_mib | 268.37890625 | valid | — |
| openai → bedrock | time_to_plateau_s | not_measured | not_measured | — |
| openai → bedrock | growth_rate_mib_per_min | -1.4534577210595634 | valid | — |
| openai → bedrock | load_s | 301 | valid | — |
| openai → bedrock | shape | 0 | valid | — |
| openai → bedrock | idle_shape | -1 | valid | — |
| openai → bedrock | idle_static | 0 | valid | — |
| openai → bedrock | idle_growth_rate_mib_per_min | -14.021685221981265 | valid | — |
| openai → bedrock | throughput_rps | bound_not_met req/s | 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 |
| openai → bedrock | throughput_rps | 685 req/s | concurrency 1; samples 4114 | 5 |
| openai → bedrock | throughput_rps | 696 req/s | concurrency 2; samples 4177 | 10 |
| openai → bedrock | throughput_rps | 696 req/s | concurrency 2; samples 4177 | 50 |
| openai → bedrock | throughput_rps | 696 req/s | concurrency 2; samples 4177 | 100 |
| openai → bedrock | throughput_rps | 696 req/s | concurrency 2; samples 4177 | unbounded |
| openai-responses → gemini | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| anthropic → gemini | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| gemini → gemini | added_latency_p50_us | 1145 | valid; gateway_c1_samples=4389; direct_c1_samples=249981 | — |
| gemini → gemini | added_latency_p99_us | 4018 | valid; gateway_c1_samples=4389; direct_c1_samples=249981 | — |
| gemini → gemini | gateway_c1_p99_us | 4054 | valid; gateway_c1_samples=4389 | — |
| gemini → gemini | direct_c1_p99_us | 36 | valid; direct_c1_samples=249981 | — |
| gemini → gemini | cpu_us_per_request | 1361.8368962787015 | valid | — |
| gemini → gemini | rps_per_cpu_second | 734.3023255813954 | valid | — |
| gemini → gemini | cost_window_conc | 8 | valid | — |
| gemini → gemini | cost_core_utilisation | 0.2855655650654452 | valid | — |
| gemini → gemini | cost_window_ok | 5052 | valid | — |
| gemini → gemini | cost_window_rps | 841 | valid | — |
| gemini → gemini | cost_threads | 22 | valid | — |
| gemini → gemini | cost_nonvol_ctxt_per_request | 0.007719714964370546 | valid | — |
| gemini → gemini | cost_majflt | 0 | valid | — |
| gemini → gemini | added_ttft_p50_us | untestable | untestable | — |
| gemini → gemini | added_ttft_p99_us | untestable | untestable | — |
| gemini → gemini | added_gap_p50_us | untestable | untestable | — |
| gemini → gemini | added_gap_p99_us | untestable | untestable | — |
| gemini → gemini | streams_sustained_fps | untestable | untestable | — |
| gemini → gemini | streams_sustained | untestable | untestable | — |
| gemini → gemini | idle_rss_mib | 124.1015625 | valid | — |
| gemini → gemini | steady_state_rss_mib | 245.73828125 | valid | — |
| gemini → gemini | recovered_rss_mib | 139.9609375 | valid | — |
| gemini → gemini | peak_rss_mib | 249.89453125 | valid | — |
| gemini → gemini | peak_rss_hwm_mib | 263.9375 | valid | — |
| gemini → gemini | time_to_plateau_s | not_measured | not_measured | — |
| gemini → gemini | growth_rate_mib_per_min | -0.8131857410506222 | valid | — |
| gemini → gemini | load_s | 301 | valid | — |
| gemini → gemini | shape | 0 | valid | — |
| gemini → gemini | idle_shape | -1 | valid | — |
| gemini → gemini | idle_static | 0 | valid | — |
| gemini → gemini | idle_growth_rate_mib_per_min | -14.014769680173412 | valid | — |
| gemini → gemini | throughput_rps | bound_not_met req/s | 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 |
| gemini → gemini | throughput_rps | 865 req/s | concurrency 2; samples 5193 | 5 |
| gemini → gemini | throughput_rps | 868 req/s | concurrency 4; samples 5212 | 10 |
| gemini → gemini | throughput_rps | 868 req/s | concurrency 4; samples 5212 | 50 |
| gemini → gemini | throughput_rps | 868 req/s | concurrency 4; samples 5212 | 100 |
| gemini → gemini | throughput_rps | 868 req/s | concurrency 4; samples 5212 | unbounded |
| openai → gemini | added_latency_p50_us | 1199 | valid; gateway_c1_samples=4241; direct_c1_samples=249376 | — |
| openai → gemini | added_latency_p99_us | 3969 | valid; gateway_c1_samples=4241; direct_c1_samples=249376 | — |
| openai → gemini | gateway_c1_p99_us | 4005 | valid; gateway_c1_samples=4241 | — |
| openai → gemini | direct_c1_p99_us | 36 | valid; direct_c1_samples=249376 | — |
| openai → gemini | cpu_us_per_request | 1420.5373880441575 | valid | — |
| openai → gemini | rps_per_cpu_second | 703.9589442815249 | valid | — |
| openai → gemini | cost_window_conc | 8 | valid | — |
| openai → gemini | cost_core_utilisation | 0.2829118072434173 | valid | — |
| openai → gemini | cost_window_ok | 4801 | valid | — |
| openai → gemini | cost_window_rps | 800 | valid | — |
| openai → gemini | cost_threads | 22 | valid | — |
| openai → gemini | cost_nonvol_ctxt_per_request | 0.006040408248281608 | valid | — |
| openai → gemini | cost_majflt | 0 | valid | — |
| openai → gemini | added_ttft_p50_us | untestable | untestable | — |
| openai → gemini | added_ttft_p99_us | untestable | untestable | — |
| openai → gemini | added_gap_p50_us | untestable | untestable | — |
| openai → gemini | added_gap_p99_us | untestable | untestable | — |
| openai → gemini | streams_sustained_fps | untestable | untestable | — |
| openai → gemini | streams_sustained | untestable | untestable | — |
| openai → gemini | idle_rss_mib | 125.1796875 | valid | — |
| openai → gemini | steady_state_rss_mib | 247.013671875 | valid | — |
| openai → gemini | recovered_rss_mib | 142.2265625 | valid | — |
| openai → gemini | peak_rss_mib | 250.59765625 | valid | — |
| openai → gemini | peak_rss_hwm_mib | 265.39453125 | valid | — |
| openai → gemini | time_to_plateau_s | not_measured | not_measured | — |
| openai → gemini | growth_rate_mib_per_min | -1.3601061120886508 | valid | — |
| openai → gemini | load_s | 301 | valid | — |
| openai → gemini | shape | 0 | valid | — |
| openai → gemini | idle_shape | -1 | valid | — |
| openai → gemini | idle_static | 0 | valid | — |
| openai → gemini | idle_growth_rate_mib_per_min | -14.117312549343774 | valid | — |
| openai → gemini | throughput_rps | bound_not_met req/s | 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 |
| openai → gemini | throughput_rps | 800 req/s | concurrency 1; samples 4805 | 5 |
| openai → gemini | throughput_rps | 809 req/s | concurrency 2; samples 4857 | 10 |
| openai → gemini | throughput_rps | 809 req/s | concurrency 2; samples 4857 | 50 |
| openai → gemini | throughput_rps | 809 req/s | concurrency 2; samples 4857 | 100 |
| openai → gemini | throughput_rps | 809 req/s | concurrency 2; samples 4857 | unbounded |
| bedrock → gemini | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| cohere → gemini | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| anthropic → openai-responses | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| openai-responses → openai-responses | added_latency_p50_us | 1238 | valid; gateway_c1_samples=4169; direct_c1_samples=265251 | — |
| openai-responses → openai-responses | added_latency_p99_us | 3854 | valid; gateway_c1_samples=4169; direct_c1_samples=265251 | — |
| openai-responses → openai-responses | gateway_c1_p99_us | 3881 | valid; gateway_c1_samples=4169 | — |
| openai-responses → openai-responses | direct_c1_p99_us | 27 | valid; direct_c1_samples=265251 | — |
| openai-responses → openai-responses | cpu_us_per_request | 1427.6795005202914 | valid | — |
| openai-responses → openai-responses | rps_per_cpu_second | 700.4373177842565 | valid | — |
| openai-responses → openai-responses | cost_window_conc | 8 | valid | — |
| openai-responses → openai-responses | cost_core_utilisation | 0.2847080853636584 | valid | — |
| openai-responses → openai-responses | cost_window_ok | 4805 | valid | — |
| openai-responses → openai-responses | cost_window_rps | 800 | valid | — |
| openai-responses → openai-responses | cost_threads | 22 | valid | — |
| openai-responses → openai-responses | cost_nonvol_ctxt_per_request | 0.004162330905306972 | valid | — |
| openai-responses → openai-responses | cost_majflt | 0 | valid | — |
| openai-responses → openai-responses | added_ttft_p50_us | untestable | untestable | — |
| openai-responses → openai-responses | added_ttft_p99_us | untestable | untestable | — |
| openai-responses → openai-responses | added_gap_p50_us | untestable | untestable | — |
| openai-responses → openai-responses | added_gap_p99_us | untestable | untestable | — |
| openai-responses → openai-responses | streams_sustained_fps | untestable | untestable | — |
| openai-responses → openai-responses | streams_sustained | untestable | untestable | — |
| openai-responses → openai-responses | idle_rss_mib | 125.03125 | valid | — |
| openai-responses → openai-responses | steady_state_rss_mib | 243.68359375 | valid | — |
| openai-responses → openai-responses | recovered_rss_mib | 141.9609375 | valid | — |
| openai-responses → openai-responses | peak_rss_mib | 249.5 | valid | — |
| openai-responses → openai-responses | peak_rss_hwm_mib | 264.3125 | valid | — |
| openai-responses → openai-responses | time_to_plateau_s | not_measured | not_measured | — |
| openai-responses → openai-responses | growth_rate_mib_per_min | -1.9357944151524555 | valid | — |
| openai-responses → openai-responses | load_s | 301 | valid | — |
| openai-responses → openai-responses | shape | 0 | valid | — |
| openai-responses → openai-responses | idle_shape | -1 | valid | — |
| openai-responses → openai-responses | idle_static | 0 | valid | — |
| openai-responses → openai-responses | idle_growth_rate_mib_per_min | -14.082313638424981 | valid | — |
| openai-responses → openai-responses | throughput_rps | bound_not_met req/s | 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 |
| openai-responses → openai-responses | throughput_rps | 809 req/s | concurrency 2; samples 4855 | 5 |
| openai-responses → openai-responses | throughput_rps | 809 req/s | concurrency 2; samples 4855 | 10 |
| openai-responses → openai-responses | throughput_rps | 809 req/s | concurrency 2; samples 4855 | 50 |
| openai-responses → openai-responses | throughput_rps | 809 req/s | concurrency 2; samples 4855 | 100 |
| openai-responses → openai-responses | throughput_rps | 809 req/s | concurrency 2; samples 4855 | unbounded |
| openai → openai-responses | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| gemini → openai-responses | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| bedrock → openai-responses | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |
| cohere → openai-responses | — | — | not_declared: Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-preflight.json | — |

## 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 |
| --- | --- | --- | --- | --- | --- | --- | --- |
| bedrock → openai | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| gemini → openai | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| openai → openai | served | — | — | — | Recorded run evidence | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| openai-responses → openai | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| anthropic → openai | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| cohere → openai | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| gemini → cohere | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| cohere → cohere | served | — | — | — | Recorded run evidence | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| bedrock → cohere | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| openai → cohere | served | — | — | — | Recorded run evidence | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| openai-responses → cohere | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| anthropic → cohere | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| gemini → anthropic | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| cohere → anthropic | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| bedrock → anthropic | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| anthropic → anthropic | served | — | — | — | Recorded run evidence | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| openai → anthropic | served | — | — | — | Recorded run evidence | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| openai-responses → anthropic | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| gemini → bedrock | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| bedrock → bedrock | served | — | — | — | Recorded run evidence | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| openai-responses → bedrock | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| anthropic → bedrock | served | — | — | — | Recorded run evidence | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| cohere → bedrock | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| openai → bedrock | served | — | — | — | Recorded run evidence | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| openai-responses → gemini | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| anthropic → gemini | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| gemini → gemini | served | — | — | — | Recorded run evidence | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| openai → gemini | served | — | — | — | Recorded run evidence | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| bedrock → gemini | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| cohere → gemini | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| anthropic → openai-responses | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| openai-responses → openai-responses | served | — | — | — | Recorded run evidence | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| openai → openai-responses | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| gemini → openai-responses | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| bedrock → openai-responses | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |
| cohere → openai-responses | not_declared | — | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 1](https://onthebench.ai/#note-1) | [recorded note 2](https://onthebench.ai/#note-2) | — | portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3 |

**Harness commit**: `d240badd966933fa19eec0606b2d1f3945f18ad2`

**Mock SHA-256**: `99d01f709d8d9c1f49055aa265d5eb7475264d6b513911fe317aee74a71190f8`

**Load generator SHA-256**: `not recorded`

**somaxconn**: 65535

**tcp_max_syn_backlog**: 65535

## 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 → openai | 1 | {"200":1} | 0 |
| openai → openai | 2 | {"200":2} | 0 |
| openai → openai | 4 | {"200":4} | 0 |
| openai → openai | 8 | {"200":8} | 0 |
| openai → openai | 16 | {"200":16} | 0 |
| openai → openai | 32 | {"200":32} | 0 |
| openai → openai | 64 | {"200":64} | 0 |
| openai → openai | 128 | {"200":128} | 0 |
| openai → openai | 256 | {"200":256} | 0 |
| openai → openai | 512 | {"200":512} | 0 |
| openai → openai | 1024 | {"200":1024} | 0 |
| openai → openai | 2048 | {"200":1810} | 238 |
| openai → openai | 1536 | {"200":703} | 833 |
| openai → openai | 1280 | {"200":1280} | 0 |
| openai → openai | 1408 | {"200":1408} | 0 |
| openai → openai | 1472 | {"200":1472} | 0 |
| openai → openai | 1504 | {"200":1504} | 0 |
| openai → openai | 1520 | {"200":1507} | 13 |
| openai → openai | 1512 | {"200":1512} | 0 |
| openai → openai | 1516 | {"200":1516} | 0 |
| openai → openai | 1518 | {"200":1518} | 0 |
| openai → openai | 1519 | {"200":1519} | 0 |
| openai → openai | 1519 | {"200":1445} | 74 |
| openai → openai | 1519 | {"200":1519} | 0 |
| anthropic → anthropic | 1 | {"200":1} | 0 |
| anthropic → anthropic | 2 | {"200":2} | 0 |
| anthropic → anthropic | 4 | {"200":4} | 0 |
| anthropic → anthropic | 8 | {"200":8} | 0 |
| anthropic → anthropic | 16 | {"200":16} | 0 |
| anthropic → anthropic | 32 | {"200":32} | 0 |
| anthropic → anthropic | 64 | {"200":64} | 0 |
| anthropic → anthropic | 128 | {"200":128} | 0 |
| anthropic → anthropic | 256 | {"200":256} | 0 |
| anthropic → anthropic | 512 | {"200":512} | 0 |
| anthropic → anthropic | 1024 | {"200":1024} | 0 |
| anthropic → anthropic | 2048 | {"200":1798} | 250 |
| anthropic → anthropic | 1536 | {"200":1151} | 385 |
| anthropic → anthropic | 1280 | {"200":918} | 362 |
| anthropic → anthropic | 1152 | {"200":919} | 233 |
| anthropic → anthropic | 1088 | {"200":1088} | 0 |
| anthropic → anthropic | 1056 | {"200":1056} | 0 |
| anthropic → anthropic | 1072 | {"200":1072} | 0 |
| anthropic → anthropic | 1080 | {"200":1080} | 0 |
| anthropic → anthropic | 1084 | {"200":1084} | 0 |
| anthropic → anthropic | 1086 | {"200":1086} | 0 |
| anthropic → anthropic | 1087 | {"200":1087} | 0 |
| anthropic → anthropic | 1087 | {"200":1087} | 0 |
| anthropic → anthropic | 1087 | {"200":1087} | 0 |
| openai → anthropic | 1 | {"200":1} | 0 |
| openai → anthropic | 2 | {"200":2} | 0 |
| openai → anthropic | 4 | {"200":4} | 0 |
| openai → anthropic | 8 | {"200":8} | 0 |
| openai → anthropic | 16 | {"200":16} | 0 |
| openai → anthropic | 32 | {"200":32} | 0 |
| openai → anthropic | 64 | {"200":64} | 0 |
| openai → anthropic | 128 | {"200":128} | 0 |
| openai → anthropic | 256 | {"200":256} | 0 |
| openai → anthropic | 512 | {"200":512} | 0 |
| openai → anthropic | 1024 | {"200":1024} | 0 |
| openai → anthropic | 2048 | {"200":1626} | 422 |
| openai → anthropic | 1536 | {"200":1211} | 325 |
| openai → anthropic | 1280 | {"200":854} | 426 |
| openai → anthropic | 1152 | {"200":950} | 202 |
| openai → anthropic | 1088 | {"200":724} | 364 |
| openai → anthropic | 1056 | {"200":1050} | 6 |
| openai → anthropic | 1040 | {"200":1040} | 0 |
| openai → anthropic | 1048 | {"200":1048} | 0 |
| openai → anthropic | 1052 | {"200":1052} | 0 |
| openai → anthropic | 1054 | {"200":1054} | 0 |
| openai → anthropic | 1055 | {"200":1055} | 0 |
| openai → anthropic | 1055 | {"200":1055} | 0 |
| openai → anthropic | 1055 | {"200":1055} | 0 |

## Recorded notes

Text recorded with the cells above. Each note is shown once; rows refer to it by number.

- Pinned v1.15.2 source ca7712947ec8c0887803990b746740d36011c697: src/index.ts registers chat-completions, messages, responses and raw proxy routes; provider operation tables determine translations. Recorder preflight proves 11 exact protocol pairs. Raw proxy passes native requests through without cross-protocol conversion, so only its matching diagonal is enabled. OpenAI Chat Completions and Responses share a provider but remain distinct upstream operations. Evidence: results/diagnostics/refresh-configuration/portkey-protocol-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-portkey-670dce412ba19689b30f84f64cb8b9c31fc07125970bd210617247c56c23bfcc](https://onthebench.ai/gateways/runs/run-portkey-670dce412ba19689b30f84f64cb8b9c31fc07125970bd210617247c56c23bfcc)  ·  [complete raw record including RSS series](https://onthebench.ai/api/v1/runs/run-portkey-670dce412ba19689b30f84f64cb8b9c31fc07125970bd210617247c56c23bfcc/raw.json)

## Configuration recorded by the run

```
Not recorded.
```

## Compare a shared protocol cell

- [Portkey vs agentgateway](https://onthebench.ai/gateways/compare/portkey-vs-agentgateway)
- [Portkey vs AISIX (api7)](https://onthebench.ai/gateways/compare/portkey-vs-aisix)
- [Portkey vs APISIX](https://onthebench.ai/gateways/compare/portkey-vs-apisix)
- [Portkey vs Bifrost](https://onthebench.ai/gateways/compare/portkey-vs-bifrost)
- [Portkey vs Busbar](https://onthebench.ai/gateways/compare/portkey-vs-busbar)
- [Portkey vs GoModel](https://onthebench.ai/gateways/compare/portkey-vs-gomodel)
- [Portkey vs Helicone](https://onthebench.ai/gateways/compare/portkey-vs-helicone)
- [Portkey vs Higress](https://onthebench.ai/gateways/compare/portkey-vs-higress)
- [Portkey vs Kong](https://onthebench.ai/gateways/compare/portkey-vs-kong)
- [Portkey vs LiteLLM · Python](https://onthebench.ai/gateways/compare/portkey-vs-litellm-python)
- [Portkey vs LiteLLM · Rust](https://onthebench.ai/gateways/compare/portkey-vs-litellm-rust)
- [Portkey vs One-API](https://onthebench.ai/gateways/compare/portkey-vs-one-api)
- [Portkey vs Plano](https://onthebench.ai/gateways/compare/portkey-vs-plano)
- [Portkey vs TensorZero](https://onthebench.ai/gateways/compare/portkey-vs-tensorzero)

## Historical measurements

- [2026-11-10T17:39:55Z — portkeyai/gateway:1.15.2@sha256:9065e15701d5f90ae1728c57369f8cec137d8c7dffcd027a402a67e64f27b9d3](https://onthebench.ai/gateways/runs/run-portkey-670dce412ba19689b30f84f64cb8b9c31fc07125970bd210617247c56c23bfcc)
- [2026-08-03T22:17:41Z — portkeyai/gateway:1.15.2](https://onthebench.ai/gateways/runs/run-portkey-55e6c9f8627b83ac5cc64c85cff37ec1fbe3c2f4e07ec33ca56aa14ae2c8e16b)
