{ "_comment": "Gateway per-request overhead vs a local mock OpenAI upstream. Sequential, localhost, interleaved rounds; median-of-round-medians. NOT a throughput/load test. See docs/methodology.md.", "unit": "ms", "results": [ { "name": "bifrost", "version": "docker-sha256:95caedb1c368", "measuredAt": "2026-07-10T09:37:24.347Z", "env": { "runner": "github-actions", "node": "v22.23.1", "platform": "linux" }, "config": { "rounds": 7, "perRound": 25, "warmup": 15 }, "verdict": { "rounds": 7, "direct_ms": 1.8, "gateway_ms": 2.41, "overhead_ms": 0.62, "overhead_iqr_ms": [ 0.59, 0.78 ] }, "directSummary": { "n": 175, "p50": 1.8, "p90": 2.01, "p99": 2.72, "min": 0.61, "max": 4.29 }, "gatewaySummary": { "n": 175, "p50": 2.45, "p90": 3, "p99": 4.19, "min": 1.25, "max": 5.4 } }, { "name": "portkey-oss", "version": "1.15.2", "measuredAt": "2026-07-10T09:37:12.710Z", "env": { "runner": "github-actions", "node": "v22.23.1", "platform": "linux" }, "config": { "rounds": 7, "perRound": 25, "warmup": 15 }, "verdict": { "rounds": 7, "direct_ms": 1.8, "gateway_ms": 4.44, "overhead_ms": 2.65, "overhead_iqr_ms": [ 2.57, 2.83 ] }, "directSummary": { "n": 175, "p50": 1.8, "p90": 1.99, "p99": 2.74, "min": 0.65, "max": 3.36 }, "gatewaySummary": { "n": 175, "p50": 4.52, "p90": 5.05, "p99": 6.06, "min": 3.17, "max": 7.06 } }, { "name": "litellm", "version": "1.91.1", "measuredAt": "2026-07-10T09:37:05.801Z", "env": { "runner": "github-actions", "node": "v22.23.1", "platform": "linux" }, "config": { "rounds": 7, "perRound": 25, "warmup": 15 }, "verdict": { "rounds": 7, "direct_ms": 1.8, "gateway_ms": 7.54, "overhead_ms": 5.83, "overhead_iqr_ms": [ 5.55, 5.86 ] }, "directSummary": { "n": 175, "p50": 1.82, "p90": 2.01, "p99": 2.78, "min": 0.63, "max": 2.9 }, "gatewaySummary": { "n": 175, "p50": 7.58, "p90": 8.34, "p99": 9.62, "min": 6.13, "max": 11.99 } } ] }