TensorZero vs Kong
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.
Requested cell: OpenAI → OpenAI. Metric: Added latency p99 (µs).
Comparability: Comparable. Ratios require an identical protocol pair, hardware, architecture, metric definition, clean harness commit, recorded mock digest, and metric-specific workload. Missing evidence never matches missing evidence. These are observed gateway measurements against a local mock, not a claim about model latency or statistical significance.
- Load-generator binary digests were not recorded; instrument equivalence is based on the identical clean harness source commit and mock digest.
- Measurements were made at different times on independently provisioned machines.
Ratio (Kong ÷ TensorZero): 0.01. No statistical significance is asserted.
Comparison JSON · All TensorZero evidence · All Kong evidence
TensorZero
Build tensorzero/gateway:2026.6.0. Measured . Run evidence
| Ingress → upstream | Metric | Value | State / reason | p99 bound (ms) |
|---|---|---|---|---|
| OpenAI → OpenAI | Added latency p50 (µs) | 40,976 | Valid · n=150 gateway / 253,662 direct | — |
| OpenAI → OpenAI | Added latency p99 (µs) | 41,966 | Valid · n=150 gateway / 253,662 direct | — |
| OpenAI → OpenAI | Gateway p99 at concurrency 1 (µs) | 41,996 | Valid · n=150 | — |
| OpenAI → OpenAI | Direct-to-mock p99 at concurrency 1 (µs) | 30 | Valid · n=253,662 | — |
| OpenAI → OpenAI | CPU per request (µs) | 459.1 | Valid | — |
| OpenAI → OpenAI | Requests per CPU-second | 2,178.2 | Valid | — |
| OpenAI → OpenAI | Cost window concurrency | 8 | Valid | — |
| OpenAI → OpenAI | Cost window core utilisation | 0.023 | Valid | — |
| OpenAI → OpenAI | Cost window successful requests | 1,198 | Valid | — |
| OpenAI → OpenAI | Cost window throughput (req/s) | 199 | Valid | — |
| OpenAI → OpenAI | Cost window threads | 6 | Valid | — |
| OpenAI → OpenAI | Non-voluntary context switches per request | 0 | Valid | — |
| OpenAI → OpenAI | Cost window major page faults | 0 | Valid | — |
| OpenAI → OpenAI | Added time to first event p50 (µs) | 707 | Valid · n=100 gateway / 100 direct | — |
| OpenAI → OpenAI | Added time to first event p99 (µs) | 746 | 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) | — | Below resolution · n=63 gateway / 63 direct | — |
| OpenAI → OpenAI | Streams sustained (frames/s) | 8,048.6 | Valid | — |
| OpenAI → OpenAI | Streams sustained | 1,991 | Valid | — |
| OpenAI → OpenAI | Idle RSS (MiB) | 48.75 | Valid | — |
| OpenAI → OpenAI | Steady-state RSS (MiB) | 67.62 | Valid | — |
| OpenAI → OpenAI | Recovered RSS (MiB) | 67.62 | Valid | — |
| OpenAI → OpenAI | Peak RSS (MiB) | 67.62 | Valid | — |
| OpenAI → OpenAI | Peak RSS high-water mark (MiB) | 67.62 | Valid | — |
| OpenAI → OpenAI | Time to plateau (s) | 60 | 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) | 0.025 | 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) | — | 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 | Throughput (req/s) | 6,275 | concurrency 256; n=37,658 | 50 ms |
| OpenAI → OpenAI | Throughput (req/s) | 11,879 | concurrency 512; n=71,288 | 100 ms |
| OpenAI → OpenAI | Throughput (req/s) | 12,324 | concurrency 1,024; n=73,956 | Unbounded |
Kong
Build kong:3.9.4. Measured . Run evidence
| Ingress → upstream | Metric | Value | State / reason | p99 bound (ms) |
|---|---|---|---|---|
| OpenAI → OpenAI | Added latency p50 (µs) | 213 | Valid · n=23,861 gateway / 246,038 direct | — |
| OpenAI → OpenAI | Added latency p99 (µs) | 410 | Valid · n=23,861 gateway / 246,038 direct | — |
| OpenAI → OpenAI | Gateway p99 at concurrency 1 (µs) | 440 | Valid · n=23,861 | — |
| OpenAI → OpenAI | Direct-to-mock p99 at concurrency 1 (µs) | 30 | Valid · n=246,038 | — |
| OpenAI → OpenAI | CPU per request (µs) | 220.7 | Valid | — |
| OpenAI → OpenAI | Requests per CPU-second | 4,531.3 | Valid | — |
| OpenAI → OpenAI | Cost window concurrency | 8 | Valid | — |
| OpenAI → OpenAI | Cost window core utilisation | 0.787 | Valid | — |
| OpenAI → OpenAI | Cost window successful requests | 85,868 | Valid | — |
| OpenAI → OpenAI | Cost window throughput (req/s) | 14,309 | Valid | — |
| OpenAI → OpenAI | Cost window threads | 5 | Valid | — |
| OpenAI → OpenAI | Non-voluntary context switches per request | 0.000734 | Valid | — |
| OpenAI → OpenAI | Cost window major page faults | 0 | Valid | — |
| OpenAI → OpenAI | Added time to first event p50 (µs) | 106,009 | Valid · n=100 gateway / 100 direct | — |
| OpenAI → OpenAI | Added time to first event p99 (µs) | 106,433 | 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) | 168,648 | Valid · n=63 gateway / 63 direct | — |
| OpenAI → OpenAI | Streams sustained (frames/s) | 3,731.9 | Valid | — |
| OpenAI → OpenAI | Streams sustained | 78 | Valid | — |
| OpenAI → OpenAI | Idle RSS (MiB) | 406.2 | Valid | — |
| OpenAI → OpenAI | Steady-state RSS (MiB) | 617.4 | Valid | — |
| OpenAI → OpenAI | Recovered RSS (MiB) | 617.5 | Valid | — |
| OpenAI → OpenAI | Peak RSS (MiB) | 617.7 | Valid | — |
| OpenAI → OpenAI | Peak RSS high-water mark (MiB) | 618.6 | Valid | — |
| OpenAI → OpenAI | Time to plateau (s) | 66 | Valid | — |
| OpenAI → OpenAI | Memory growth (MiB/min) | 0.105 | Valid | — |
| OpenAI → OpenAI | Memory load duration (s) | 300 | 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) | 0.241 | Valid | — |
| OpenAI → OpenAI | Throughput (req/s) | 13,315 | concurrency 4; n=79,905 | 1 ms |
| OpenAI → OpenAI | Throughput (req/s) | 19,190 | concurrency 8; n=115,157 | 5 ms |
| OpenAI → OpenAI | Throughput (req/s) | 21,287 | concurrency 64; n=127,725 | 10 ms |
| OpenAI → OpenAI | Throughput (req/s) | 22,171 | concurrency 128; n=133,046 | 50 ms |
| OpenAI → OpenAI | Throughput (req/s) | 22,171 | concurrency 128; n=133,046 | 100 ms |
| OpenAI → OpenAI | Throughput (req/s) | 22,171 | concurrency 128; n=133,046 | Unbounded |