CAP
Reasoning
League table — best at reasoning · 30
mean percentile across this capability's benchmarks| # | Model | League | Coverage | $/1M | tok/s |
|---|---|---|---|---|---|
| 1 | GPT-5.6 Sol (max) | 99.8 | 2/5 | $11.25 | 85 |
| 2 | Claude Fable 5 | 99.8 | 2/5 | $20 | 65 |
| 3 | Gemini 3.1 Pro Preview | 99.5 | 2/5 | $4.5 | 140 |
| 4 | Claude Opus 4.8 | 99.5 | 2/5 | $10 | 59 |
| 5 | GPT-5.6 Sol | 99.3 | 2/5 | $11.25 | 71 |
| 6 | GPT-5.5 | 99.2 | 2/5 | $11.25 | 72 |
| 7 | Muse Spark 1.1 | 99.0 | 2/5 | $2 | 129 |
| 8 | GPT-5.6 Terra (max) | 98.9 | 2/5 | $5.63 | 166 |
| 9 | GPT-5.4 | 98.7 | 2/5 | $5.63 | 161 |
| 10 | Grok 4.5 | 98.7 | 2/5 | $3 | 110 |
| 11 | Gemini 3.5 Flash | 98.6 | 2/5 | $3.38 | 238 |
| 12 | GPT-5.6 Terra | 98.0 | 2/5 | $5.63 | 130 |
| 13 | GPT-5.3 Codex | 97.9 | 2/5 | $4.81 | 99 |
| 14 | GLM-5.2 (max) | 97.9 | 2/5 | $2.15 | 200 |
| 15 | Claude Opus 4.7 | 97.8 | 2/5 | $10 | 52 |
| 16 | Qwen3.7 Max | 97.7 | 2/5 | $3.75 | 197 |
| 17 | Claude Sonnet 5 | 97.6 | 2/5 | $4 | 79 |
| 18 | Muse Spark | 97.4 | 2/5 | — | — |
| 19 | MiniMax-M3 | 97.4 | 2/5 | $0.525 | 111 |
| 20 | Gemini 3 Pro Preview | 97.3 | 2/5 | $4.5 | — |
| 21 | GPT-5.6 Luna (max) | 97.2 | 2/5 | $2.25 | 234 |
| 22 | Kimi K2.6 | 96.7 | 2/5 | $1.71 | 43 |
| 23 | Claude Opus 4.6 | 96.7 | 2/5 | $10 | 48 |
| 24 | Grok Build 0.1 0616 | 96.5 | 2/5 | $1.25 | 59 |
| 25 | o3-pro | 96.3 | 1/5 | $35 | 26 |
| 26 | GPT-5.2 | 96.2 | 2/5 | $4.81 | 81 |
| 27 | GPT-5.6 Luna | 96.1 | 2/5 | $2.25 | 222 |
| 28 | DeepSeek V4 Pro | 96.1 | 2/5 | $0.544 | 64 |
| 29 | Grok 4.3 | 96.1 | 2/5 | $1.56 | 152 |
| 30 | Gemini 3 Flash Preview | 95.9 | 2/5 | $1.13 | 202 |
League = a model's mean percentile across the 5 benchmarksin this capability (best score per benchmark; partial coverage shown). Price & speed via Artificial Analysis.
Benchmarks in this capability (5)
ARC-AGI0 scores
BIG-Bench Hard1480 scores
GPQA Diamond1909 scores
Humanity's Last Exam404 scores
MuSR1480 scores