Hi everyone ,
Its very important for us to understand the timing performance for the requests we do through ai studio apis
Currently we experience sporadic latency from gemini 3.1 pro that has been there for months and it messes up with out internal timeout fallbacks as there are no metrics available to predict what is a good timeout we can set internally
Metrics that I suggest to add , compared to openrouter logs
Performance & Timing Metrics
| Metric | OpenRouter Metadata | Google AI Studio Logs |
|---|---|---|
| TTFT (Time to First Token) | Yes. Logged for every request; critical for assessing UI responsiveness. | No. Currently a high-priority feature request; not natively in logs. |
| TPS (Tokens Per Second) | Yes. Calculated and logged as “throughput” metadata. | No. You must manually calculate this from the response timestamp. |
| Wall-Clock Latency | Yes. Total E2E time including routing overhead. | Minimal. Usually limited to a basic request timestamp. |
| Provider Latency | Yes. Shows exactly how much time the underlying provider took vs. the router. | N/A. Google is the only provider; internal network latency is hidden. |
Thank you
Jay