LMArena Agent Arena Code Steering Burden
Agents · 2026-08-14
Signed percentage-point Agent Arena Code category diagnostic signal for the steering_burden signal (displayed by the source as Steerability), measuring response to user corrections, recomputed on sessions tagged as writing or debugging code, automating workflows or analyzing data. 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 | 22.7 |
| Claude Fable 5.1 | 20.4 |
| GPT-6.1 Sol | 17.8 |
| GPT-6 Astra | 17.8 |
| Gemini 4 Argon | 16.2 |
| Claude Fable 5 | 16.0 |
| GPT-6 Sol | 16.0 |
| Claude Sonnet 5.5 | 15.9 |
| Muse Wonder Full | 15.0 |
| Claude Haiku 5.5 | 11.9 |
| GPT-5.6 Sol | 11.2 |
| GPT-5.4 | 9.7 |
| Claude Opus 5 | 9.2 |
| GPT-6 Luna | 8.8 |
| Sonnet 5 | 7.9 |
| DeepSeek V4.1 Flash 1017 | 7.1 |
| Muse Spark 1.3 | 6.9 |
| MiMo-V2.6-Flash | 6.1 |
| GPT-5.5 | 5.5 |
| GPT-5.6 Terra | 4.7 |
| Muse Wonder | 4.2 |
| GLM-5.2 | 3.9 |
| Step 5 Preview | 3.1 |
| DeepSeek V4.1 Flash | 2.9 |
| MiMo-V2.6-Pro | 2.6 |
| Hy4 Preview | 2.5 |
| Qwen3.8-Max | 1.4 |
| Kimi K3 | 1.3 |
| Grok 4.6 | 1.2 |
| Grok 4.7 | 1.1 |
| GLM-5.3 | 1.0 |
| Mistral Large 4 Preview | 0.8 |
| Hy3 | -0.1 |
| Qwen3.8-Flash-Next | -0.4 |
| DeepSeek V4 Pro 0813 | -0.9 |
| GPT-5.6 Luna | -1.3 |
| Gemini 3.8 Flash | -1.6 |
| GLM-5.3 Flash | -4.3 |
| Qwen3.8 27B | -4.6 |
| Gemini 3.7 Flash | -8.3 |