LMArena Agent Arena Work Confirmed Success
Agents · 2026-08-14
Signed percentage-point Agent Arena Work category diagnostic signal for sessions where the user explicitly confirmed task completion, 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 Sonnet 5.5 | 21.7 |
| Claude Opus 5.5 | 19.0 |
| Claude Fable 5.1 | 13.9 |
| GPT-6.1 Sol | 13.3 |
| GPT-6 Astra | 12.4 |
| Gemini 4 Argon | 12.0 |
| Muse Wonder Full | 11.8 |
| Claude Opus 5 | 11.0 |
| DeepSeek V4.1 Flash 1017 | 9.3 |
| Muse Wonder | 8.6 |
| MiMo-V2.6-Flash | 7.6 |
| GPT-6 Sol | 7.4 |
| MiMo-V2.6-Pro | 7.3 |
| Kimi K3 | 7.0 |
| DeepSeek V4.1 Flash | 6.3 |
| GLM-5.2 | 6.1 |
| Hy4 Preview | 5.9 |
| Muse Spark 1.3 | 5.6 |
| Grok 4.7 | 4.7 |
| GLM-5.3 | 4.0 |
| Qwen3.8-Max | 3.3 |
| Sonnet 5 | 2.5 |
| Claude Fable 5 | 2.5 |
| Qwen3.8-Flash-Next | 2.3 |
| GLM-5.3 Flash | 2.2 |
| GPT-5.6 Sol | 2.1 |
| Claude Haiku 5.5 | 1.9 |
| DeepSeek V4 Pro 0813 | 1.5 |
| GPT-6 Luna | 1.5 |
| Step 5 Preview | 0.8 |
| Gemini 3.8 Flash | 0.7 |
| Mistral Large 4 Preview | -0.3 |
| GPT-5.5 | -0.7 |
| Qwen3.8 27B | -1.0 |
| Grok 4.6 | -2.6 |
| GPT-5.6 Terra | -3.2 |
| GPT-5.4 | -3.8 |
| GPT-5.6 Luna | -4.2 |
| Gemini 3.7 Flash | -5.2 |
| MiMo-V2.5-Pro | -7.2 |