Benchmark
Arena Agent Bash Reliability
Agentic benchmarks measure how well models carry out tasks that take several steps. Higher scores are better.
All agentic benchmarksScores from LMArena data.
- Models measured
- 62
- Scores recorded
- 68
- Sources
- 1
- IQ weight
- 1.47
Score distribution
- -0.001
- Mean
- 0.015
- Median
- 0.105
- Std dev
- -0.129
- 10th percentile
- 0.105
- 90th percentile
- -0.491
- Min
This benchmark has a weight of 1.47 for chat models, measured over 62 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 51 | Kimi K3Max· LMArena data | -0.0561 |
| 52 | MiMo V2.6 Flash· LMArena data | -0.0568 |
| 53 | Gemini 4 CarbonHigh· LMArena data | -0.0575 |
| 54 | Gemini 4 ArgonHigh· LMArena data | -0.0596 |
| 55 | Gemini 4 NeonHigh· LMArena data | -0.0597 |
| 56 | MiniMax M3.1 Flash· LMArena data | -0.0634 |
| 57 | GLM 5.3Max· LMArena data | -0.0671 |
| 58 | Mistral Large 4· LMArena data | -0.0795 |
| 59 | Mistral Medium 3.5· LMArena data | -0.0868 |
| 60 | Spark-X2.6· LMArena data | -0.0926 |
| 61 | GLM 5.3 Flash· LMArena data | -0.129 |
| 62 | Step 5 Preview· LMArena data | -0.131 |
| 63 | Qwen 3.8 27B· LMArena data | -0.133 |
| 64 | NVIDIA Nemotron 3.5 Super VL 120BNVFP4· LMArena data | -0.134 |
| 65 | Hy3 (Deep Infra)· LMArena data | -0.138 |
| 66 | Qwen 3.8 Flash Next· LMArena data | -0.23 |
| 67 | Nemotron 3 Ultra· LMArena data | -0.252 |
| 68 | Solar Pro 4· LMArena data | -0.491 |
Showing 51–68 of 68 results.
Data credits: LMArena, CC BY 4.0 (licence)