LMArena Agent Arena Work Steering Burden
Agents · 2026-08-14
Signed percentage-point Agent Arena Work category diagnostic signal for the steering_burden signal (displayed by the source as Steerability), measuring response to user corrections, recomputed on sessions tagged as office work such as document creation, professional research, planning and professional writing. Values may be negative relative to the category baseline; reported confidence intervals are not standard errors.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 18.5 |
| Claude Fable 5.1 | 17.2 |
| Gemini 4 Argon | 15.3 |
| GPT-6.1 Sol | 14.6 |
| GPT-6 Astra | 14.4 |
| GPT-6 Sol | 13.9 |
| Claude Fable 5 | 13.1 |
| Claude Sonnet 5.5 | 12.2 |
| Muse Wonder Full | 11.5 |
| GPT-5.6 Sol | 9.3 |
| GPT-5.4 | 8.4 |
| Sonnet 5 | 8.2 |
| Claude Haiku 5.5 | 7.8 |
| GPT-6 Luna | 7.0 |
| Claude Opus 5 | 6.7 |
| Muse Spark 1.3 | 5.8 |
| MiMo-V2.6-Flash | 5.7 |
| DeepSeek V4.1 Flash 1017 | 4.8 |
| Muse Wonder | 4.5 |
| GPT-5.5 | 3.5 |
| GPT-5.6 Terra | 3.1 |
| Step 5 Preview | 2.8 |
| MiMo-V2.6-Pro | 2.8 |
| GLM-5.2 | 2.4 |
| DeepSeek V4.1 Flash | 2.0 |
| Qwen3.8-Max | 1.8 |
| Hy4 Preview | 1.8 |
| Kimi K3 | 1.6 |
| Hy3 | 0.6 |
| Grok 4.7 | 0.4 |
| GLM-5.3 | 0.3 |
| Mistral Large 4 Preview | 0.2 |
| Gemini 3.8 Flash | 0.2 |
| Grok 4.6 | -0.2 |
| DeepSeek V4 Pro 0813 | -0.2 |
| Qwen3.8-Flash-Next | -1.5 |
| GPT-5.6 Luna | -1.7 |
| GLM-5.3 Flash | -3.3 |
| Qwen3.8 27B | -3.6 |
| Gemini 3.7 Flash | -5.9 |