LMArena Agent Arena Steerability
Agents · 2026-06-04
Signed percentage-point Agent Arena diagnostic signal measuring how well the agent follows user steering and course corrections. Values may be negative when a model is below the leaderboard baseline; higher values indicate better controllability in interactive tasks.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Sol | 14.5 |
| Claude Fable 5 | 11.4 |
| Claude Opus 5 | 10.7 |
| Opus 4.8 | 9.6 |
| Claude Fable 5.1 | 8.4 |
| Sonnet 5 | 7.5 |
| GPT-5.5 | 6.9 |
| GPT-5.6 Sol | 6.5 |
| Gemini 3.8 Flash | 4.3 |
| GLM-5.2 | 4.2 |
| GPT-5.6 Terra | 4.1 |
| Grok 4.5 | 3.9 |
| Grok 4.6 | 3.3 |
| GLM-5.3 | 2.8 |
| GPT-5.4 | 2.4 |
| Qwen3.8-Max | 2.0 |
| GPT-5.6 Luna | 1.9 |
| Muse Spark 1.3 | 1.6 |
| DeepSeek V4 Pro 0813 | 0.5 |
| Grok 4.7 | 0.5 |
| Gemini 3.7 Flash | -0.0 |
| Kimi K3 | -0.5 |
| Hy4 Preview | -0.7 |
| Qwen3.8 27B | -1.0 |
| GLM-5.3 Flash | -1.1 |
| Hy3 | -1.3 |
| GPT-6 Astra | -1.4 |
| DeepSeek-V4-Pro | -2.8 |
| DeepSeek V4.1 Flash | -3.3 |
| Qwen3.8-Flash-Next | -3.4 |
| Qwen3.7-Max (source unspecified) | -3.5 |
| Gemini 3.1 Pro Preview | -4.0 |
| Gemini 3.6 Flash | -4.0 |
| MiMo-V2.5-Pro | -4.5 |
| Muse Spark 1.2 | -5.4 |
| Muse Spark 1.1 | -5.6 |
| MiniMax M3 | -7.0 |
| Qwen3.7-Plus (source unspecified) | -7.5 |
| Solar Pro 4 | -10.3 |
| Inkling-Small | -11.5 |