LMArena Agent Arena Work Praise vs Complaint
Agents · 2026-08-14
Signed percentage-point Agent Arena Work category diagnostic signal comparing positive user feedback against complaints, 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 | 43.6 |
| GPT-6.1 Sol | 30.9 |
| GPT-6 Astra | 28.9 |
| Claude Fable 5.1 | 26.9 |
| Gemini 4 Argon | 25.1 |
| Kimi K3.1 Pro | 24.6 |
| Claude Sonnet 5.5 | 20.6 |
| Muse Wonder Full | 16.5 |
| Claude Fable 5 | 15.3 |
| GPT-6 Sol | 13.8 |
| Claude Opus 5 | 11.6 |
| GPT-5.6 Sol | 11.5 |
| DeepSeek V4.1 Flash 1017 | 9.3 |
| Sonnet 5 | 6.5 |
| GLM-5.2 | 6.1 |
| GLM-5.3 | 5.2 |
| MiMo-V2.6-Pro | 5.0 |
| MiMo-V2.6-Flash | 3.9 |
| DeepSeek V4.1 Flash | 3.7 |
| Gemini 3.8 Flash | 3.3 |
| GPT-5.6 Terra | 3.2 |
| Hy4 Preview | 2.9 |
| Grok 4.7 | 2.8 |
| Muse Spark 1.3 | 2.7 |
| Kimi K3 | 1.9 |
| Qwen3.8-Max | 1.6 |
| GPT-6 Luna | 1.4 |
| Muse Wonder | 0.9 |
| Step 5 Preview | -0.3 |
| DeepSeek V4 Pro 0813 | -0.4 |
| Grok 4.6 | -1.5 |
| Claude Haiku 5.5 | -2.5 |
| GLM-5.3 Flash | -2.9 |
| Mistral Large 4 Preview | -3.0 |
| Qwen3.8-Flash-Next | -3.4 |
| Gemini 3.7 Flash | -3.4 |
| Gemini 3.1 Pro Preview | -4.2 |
| Qwen3.8 27B | -5.8 |
| Hy3 | -7.3 |
| Qwen3.7-Plus (source unspecified) | -10.4 |