CAP
Agents
League table — best at agents · 30
mean percentile across this capability's benchmarks| # | Model | League | Coverage | $/1M | tok/s |
|---|---|---|---|---|---|
| 1 | Claude Fable 5 | 98.9 | 2/3 | $20 | 58 |
| 2 | GLM-5.2 (max) | 97.8 | 2/3 | $2.15 | 157 |
| 3 | Gemini 3.1 Pro Preview | 95.9 | 2/3 | $4.5 | 132 |
| 4 | Claude Opus 4.8 | 95.0 | 2/3 | $10 | 63 |
| 5 | DeepSeek V4 Pro | 95.0 | 2/3 | $0.544 | 71 |
| 6 | Qwen3.6 Plus | 94.6 | 2/3 | $1.13 | 53 |
| 7 | GLM-5 | 94.6 | 2/3 | $1.55 | — |
| 8 | GPT-5.5 | 94.4 | 2/3 | $11.25 | — |
| 9 | GLM-5.1 | 94.3 | 2/3 | $2.14 | — |
| 10 | Qwen3.6 Max Preview | 94.2 | 2/3 | $2.92 | — |
| 11 | Qwen3.7 Max | 94.1 | 2/3 | $3.75 | 200 |
| 12 | Kimi K2.6 | 93.4 | 2/3 | $1.71 | — |
| 13 | Grok 4.20 0309 | 92.6 | 2/3 | $3 | — |
| 14 | Qwen3.5 397B A17B | 91.5 | 2/3 | $1.35 | 66 |
| 15 | Grok 4.3 | 91.5 | 2/3 | $1.56 | — |
| 16 | Qwen3.7 Plus | 91.1 | 2/3 | $0.7 | 53 |
| 17 | MiMo-V2-Pro | 91.0 | 2/3 | — | — |
| 18 | Step 3.7 Flash | 91.0 | 2/3 | $0.438 | 398 |
| 19 | MiMo-V2.5-Pro | 90.9 | 2/3 | $0.544 | 65 |
| 20 | Gemini 3.5 Flash | 90.7 | 2/3 | $3.38 | 250 |
| 21 | Claude Opus 4.6 | 90.2 | 2/3 | $10 | — |
| 22 | GLM-5-Turbo | 89.5 | 2/3 | — | — |
| 23 | Muse Spark | 89.2 | 2/3 | — | — |
| 24 | Kimi K2.5 | 89.1 | 2/3 | $1.2 | — |
| 25 | KAT Coder Pro V2 | 89.0 | 2/3 | $0.525 | 104 |
| 26 | Claude Opus 4.5 | 88.9 | 2/3 | $10 | — |
| 27 | Claude Opus 4.7 | 88.8 | 2/3 | $10 | — |
| 28 | GLM 5V Turbo | 88.7 | 2/3 | — | — |
| 29 | GPT-5.4 | 88.7 | 2/3 | $5.63 | — |
| 30 | MiniMax-M2.5 | 88.3 | 2/3 | $0.525 | — |
League = a model's mean percentile across the 3 benchmarksin this capability (best score per benchmark; partial coverage shown). Price & speed via Artificial Analysis.
Benchmarks in this capability (7)
ALFWorld0 scores
BabyAI0 scores
SWE-bench Verified5 scores
SWE-Gym0 scores
Terminal-Bench Hard316 scores
WebArena0 scores
τ²-bench324 scores