Ranking
Median time from sending the request to the first streamed token of any kind, so thinking time counts. Top 15 of 74.
| # | Model | First token | Speed (tok/s) | Score | Cost / question |
|---|---|---|---|---|---|
| 1 | Mistral | 0.3 s | 252 | 26.2% | $0.00019 |
| 2 | Mistral | 0.3 s | 124 | 42.5% | $0.00027 |
| 3 | Nous | 0.3 s | 41 | 26.2% | $0.01 |
| 4 | MoonshotAI | 0.4 s | 136 | 65.0% | $0.006 |
| 5 | Xiaomi | 0.4 s | 77 | 62.5% | $0.00056 |
| 6 | Qwen | 0.4 s | 135 | 37.5% | $0.0037 |
| 7 | Qwen | 0.4 s | 106 | 62.5% | $0.0017 |
| 8 | Mistral | 0.5 s | 145 | 28.8% | $0.025 |
| 9 | Microsoft | 0.5 s | 103 | 41.2% | $0.00012 |
| 10 | Amazon | 0.5 s | 180 | 22.5% | $0.000087 |
| 11 | rekaai | 0.5 s | 101 | 1.2% | $0.0008 |
| 12 | Mistral | 0.5 s | 170 | 32.5% | $0.0019 |
| 13 | NVIDIA | 0.5 s | 116 | 43.8% | $0.00049 |
| 14 | Meta | 0.5 s | 220 | 77.5% | $0.0045 |
| 15 | Z.ai | 0.5 s | 83 | 67.5% | $0.00015 |