Leaderboard
Ranks with the units still attached
Capability is not speed is not price. Sort the board by the quantity you actually need. Provider-reported numbers never occupy the same column as GBU measurements.
| Endpoint | Evidence | ||||||
|---|---|---|---|---|---|---|---|
| Llama 4 70B | 64.3 | 71.5 | $0.0071 | 55 ms | 890 | 99.84% | LIVE_MEASURED |
| Llama 4 70B | 32.3 | 71.5 | $0.0064 | 140 ms | 195 | 99.75% | LIVE_MEASURED |
| Gemini 2.5 Flash | 20.2 | 79.4 | $0.013 | 120 ms | 280 | 99.97% | LIVE_MEASURED |
| Grok 3 Mini | 16.6 | 74.6 | $0.013 | 180 ms | 220 | 99.96% | LIVE_MEASURED |
| Llama 4 405B | 13.5 | 78.2 | $0.012 | 95 ms | 210 | 99.20% | REPRODUCED |
| GPT-5 Mini | 8.50 | 78.9 | $0.020 | 240 ms | 160 | 99.93% | LIVE_MEASURED |
| Llama 4 405B | 8.05 | 78.2 | $0.030 | 90 ms | 420 | 99.81% | LIVE_MEASURED |
| Llama 4 405B | 4.76 | 78.2 | $0.016 | 180 ms | 85 | 99.20% | REPRODUCED |
| DeepSeek V3.1 | 4.52 | 82.4 | $0.024 | 310 ms | 78 | 99.41% | LIVE_MEASURED |
| Llama 4 405B | 3.43 | 78.2 | $0.036 | 220 ms | 155 | 99.68% | LIVE_MEASURED |
| Llama 4 405B | 2.97 | 78.2 | $0.032 | 280 ms | 110 | 99.72% | LIVE_MEASURED |
| Qwen3-Max | 2.56 | 81.6 | $0.036 | 260 ms | 92 | 99.70% | LIVE_MEASURED |
| Gemini 2.5 Pro | 1.96 | 86.8 | $0.053 | 340 ms | 145 | 99.95% | LIVE_MEASURED |
| Mistral Large 3 | 1.32 | 80.1 | $0.068 | 350 ms | 98 | 99.63% | LIVE_MEASURED |
| Grok 4 | 0.925 | 87.3 | $0.095 | 410 ms | 138 | 99.94% | LIVE_MEASURED |
| Claude 4.1 Sonnet | 0.724 | 84.1 | $0.108 | 390 ms | 108 | 99.90% | LIVE_MEASURED |
| GPT-5 | 0.268 | 88.6 | $0.262 | 520 ms | 95 | 99.91% | LIVE_MEASURED |
| GPT-5 | 0.217 | 88.6 | $0.292 | 610 ms | 88 | 99.88% | LIVE_MEASURED |
| Grok 4 Heavy | 0.198 | 89.8 | $0.205 | 890 ms | 62 | 99.91% | LIVE_MEASURED |
| Claude 4.1 Opus | 0.104 | 87.9 | $0.519 | 680 ms | 72 | 99.86% | LIVE_MEASURED |
| o4-pro | 0.033 | 90.4 | $0.738 | 2.40 s | 48 | 99.74% | LIVE_MEASURED |