LMArena Agent Arena Chat Tool Hallucination
Agents · 2026-08-14
Signed percentage-point Agent Arena Chat diagnostic improvement signal for reductions in apparent nonexistent or inappropriate tool use, recomputed on sessions tagged as general agentic chat such as creative writing, learning, everyday research and media generation. Atlas stores the raw improvement orientation relative to the category baseline: positive values indicate more grounded tool use and higher is better. Reported confidence intervals are not standard errors.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6.1 Sol | 0.5 |
| Claude Fable 5 | 0.5 |
| GPT-5.6 Terra | 0.5 |
| Claude Sonnet 5.5 | 0.5 |
| GLM-5.3 | 0.5 |
| Kimi K3.1 Pro | 0.5 |
| Claude Opus 5 | 0.5 |
| GPT-6 Astra | 0.5 |
| Kimi K3 | 0.5 |
| Claude Fable 5.1 | 0.5 |
| Muse Spark 1.3 | 0.5 |
| Muse Wonder | 0.5 |
| Gemini 3.1 Pro Preview | 0.4 |
| Sonnet 5 | 0.4 |
| Grok 4.7 | 0.4 |
| Claude Opus 5.5 | 0.4 |
| GLM-5.2 | 0.4 |
| GPT-6 Luna | 0.4 |
| GPT-6 Sol | 0.4 |
| GLM-5.3 Flash | 0.4 |
| DeepSeek V4.1 Flash 1017 | 0.3 |
| DeepSeek V4.1 Flash | 0.3 |
| Gemini 3.7 Flash | 0.3 |
| Grok 4.6 | 0.3 |
| Muse Wonder Full | 0.3 |
| Gemini 3.8 Flash | 0.2 |
| Step 5 Preview | 0.2 |
| Gemini 4 Argon | 0.2 |
| GPT-5.6 Sol | 0.2 |
| DeepSeek V4 Pro 0813 | 0.1 |
| Hy4 Preview | 0.1 |
| Muse Spark 1.1 | 0.1 |
| Muse Spark 1.2 | -0.1 |
| Claude Haiku 5.5 | -0.2 |
| Qwen3.7-Max (source unspecified) | -0.2 |
| MiniMax M3 | -0.3 |
| Qwen3.8 27B | -0.3 |
| MiMo-V2.6-Pro | -0.4 |
| Inkling-Small | -0.5 |
| Qwen3.8-Flash-Next | -0.5 |