Dev Tools · 1h ago
Benchmark: LLM Gateway beats OpenRouter on time-to-first-token by 35%
An open-source benchmark measured time-to-first-token for LLM Gateway and OpenRouter using Claude Haiku 4.5. LLM Gateway returned the first token in 906ms cold and 814ms warm, versus OpenRouter's 1392ms and 1232ms. The test ran 450 HTTP requests with zero errors.
Meridian48 take
The benchmark is transparent and reproducible, but results from a single vantage point and model may not generalize to all production workloads.
ai-gateway-benchmarkllm-latency