Time to First Token
Median seconds until the first output token.
Time to First Token measures perceived responsiveness for streamed responses. Artificial Analysis defines it as the time from request submission to the first received token, with lower values ranking better in this app.
Test type: Hosted endpoint latency measurement. Lower values mean the response starts sooner.
Coverage
293 models have this metric.
0.24s
Current leader: NVIDIA Nemotron Nano 9B V2 (Reasoning)
Project links
Values come from Artificial Analysis performance data in the committed snapshot.
Top TTFT Models
Top models ranked by TTFT.
Leaderboard
| Rank | Model | Creator | Value | Speed | Blended Price |
|---|---|---|---|---|---|
| #1 | NVIDIA Nemotron Nano 9B V2 (Reasoning) | NVIDIA | 0.24s | 121.6 tok/s | $0.070/M |
| #2 |
