LMArena Agent Arena Praise vs Complaint
Agents · 2026-06-04
Signed percentage-point Agent Arena diagnostic signal comparing positive user feedback against complaints. Values may be negative when a model is below the leaderboard baseline; higher values indicate more favorable user reactions to the agent’s work.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 45.3 |
| GPT-6.1 Sol | 31.3 |
| GPT-6 Astra | 29.8 |
| Claude Fable 5.1 | 27.7 |
| Gemini 4 Argon | 26.6 |
| Claude Sonnet 5.5 | 23.2 |
| Muse Wonder Full | 20.0 |
| Kimi K3.1 Pro | 19.0 |
| Claude Fable 5 | 15.3 |
| GPT-6 Sol | 14.1 |
| Claude Opus 5 | 12.4 |
| GPT-5.6 Sol | 10.4 |
| DeepSeek V4.1 Flash 1017 | 6.0 |
| GLM-5.2 | 4.8 |
| GLM-5.3 | 4.8 |
| Gemini 3.8 Flash | 4.2 |
| MiMo-V2.6-Pro | 4.1 |
| Sonnet 5 | 3.8 |
| Grok 4.7 | 3.5 |
| Hy4 Preview | 3.4 |
| DeepSeek V4.1 Flash | 2.9 |
| Muse Spark 1.3 | 2.9 |
| MiMo-V2.6-Flash | 2.6 |
| GPT-5.6 Terra | 2.0 |
| Kimi K3 | 2.0 |
| Qwen3.8-Max | 1.1 |
| Muse Wonder | 1.1 |
| DeepSeek V4 Pro 0813 | 0.6 |
| GPT-6 Luna | 0.5 |
| Grok 4.6 | -1.3 |
| Gemini 3.7 Flash | -2.2 |
| Gemini 3.1 Pro Preview | -2.4 |
| GLM-5.3 Flash | -2.9 |
| Qwen3.8-Flash-Next | -3.8 |
| Claude Haiku 5.5 | -4.0 |
| Step 5 Preview | -4.8 |
| Mistral Large 4 Preview | -5.1 |
| Qwen3.8 27B | -6.5 |
| Hy3 | -7.9 |
| Qwen3.7-Max (source unspecified) | -10.2 |