LMArena Agent Arena Chat Steering Burden
Agents · 2026-08-14
Signed percentage-point Agent Arena Chat category diagnostic signal for the steering_burden signal (displayed by the source as Steerability), measuring response to user corrections, recomputed on sessions tagged as general agentic chat such as creative writing, learning, everyday research and media generation. Values may be negative relative to the category baseline; reported confidence intervals are not standard errors.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 4 Argon | 14.3 |
| Claude Opus 5.5 | 14.0 |
| GPT-6 Astra | 11.9 |
| Claude Fable 5.1 | 10.9 |
| GPT-6.1 Sol | 9.3 |
| Muse Wonder Full | 8.2 |
| Claude Sonnet 5.5 | 7.4 |
| GPT-6 Sol | 5.5 |
| Gemini 3.8 Flash | 5.4 |
| GPT-5.6 Sol | 5.2 |
| Claude Opus 5 | 4.6 |
| Kimi K3.1 Pro | 4.5 |
| Claude Fable 5 | 3.2 |
| MiMo-V2.6-Flash | 2.9 |
| GPT-6 Luna | 2.8 |
| Sonnet 5 | 2.1 |
| Muse Wonder | 2.1 |
| Muse Spark 1.3 | 1.8 |
| MiMo-V2.6-Pro | 1.7 |
| Gemini 3.7 Flash | 1.6 |
| Grok 4.6 | 1.6 |
| DeepSeek V4 Pro 0813 | 1.5 |
| Qwen3.8-Max | 1.0 |
| Mistral Large 4 Preview | 1.0 |
| DeepSeek V4.1 Flash 1017 | 0.8 |
| GPT-5.6 Terra | 0.6 |
| Grok 4.7 | 0.6 |
| Hy4 Preview | 0.3 |
| Gemini 3.1 Pro Preview | 0.1 |
| Kimi K3 | -0.2 |
| GLM-5.3 Flash | -0.3 |
| DeepSeek V4.1 Flash | -0.5 |
| Step 5 Preview | -0.9 |
| GLM-5.3 | -2.0 |
| Claude Haiku 5.5 | -2.2 |
| GLM-5.2 | -2.3 |
| Muse Spark 1.2 | -3.4 |
| Qwen3.7-Plus (source unspecified) | -3.9 |
| Qwen3.7-Max (source unspecified) | -3.9 |
| Hy3 | -4.2 |