LMArena Agent Arena Tool Hallucination
Agents · 2026-06-04
Signed percentage-point Agent Arena diagnostic improvement signal for reductions in apparent nonexistent or inappropriate tool use relative to the leaderboard baseline. Atlas stores the source's raw improvement orientation: positive values indicate more grounded tool use and higher values are better. Reported confidence intervals are not standard errors.
Top models (higher is better)
| Model | Score |
|---|---|
| Kimi K3.1 Pro | 0.7 |
| GPT-6.1 Sol | 0.7 |
| Kimi K3 | 0.7 |
| GLM-5.2 | 0.7 |
| GPT-6 Astra | 0.7 |
| Claude Opus 5 | 0.7 |
| GLM-5.3 | 0.7 |
| GLM-5.3 Flash | 0.7 |
| Grok 4.7 | 0.7 |
| Step 5 Preview | 0.7 |
| Muse Spark 1.3 | 0.7 |
| Claude Sonnet 5.5 | 0.6 |
| GPT-6 Sol | 0.6 |
| Claude Fable 5 | 0.6 |
| GPT-5.6 Sol | 0.6 |
| GPT-6 Luna | 0.6 |
| Grok 4.6 | 0.6 |
| Claude Fable 5.1 | 0.6 |
| DeepSeek V4 Pro 0813 | 0.6 |
| Claude Opus 5.5 | 0.6 |
| Gemini 3.8 Flash | 0.6 |
| DeepSeek V4.1 Flash | 0.6 |
| GPT-5.6 Terra | 0.6 |
| Gemini 3.7 Flash | 0.5 |
| Sonnet 5 | 0.4 |
| DeepSeek V4.1 Flash 1017 | 0.4 |
| Muse Wonder Full | 0.4 |
| Gemini 4 Argon | 0.4 |
| Muse Spark 1.1 | 0.3 |
| Muse Spark 1.2 | 0.2 |
| Hy4 Preview | 0.1 |
| MiMo-V2.6-Pro | -0.1 |
| Qwen3.8 27B | -0.1 |
| Gemini 3.1 Pro Preview | -0.1 |
| Qwen3.7-Max (source unspecified) | -0.2 |
| Qwen3.8-Flash-Next | -0.8 |
| Qwen3.8-Max | -0.8 |
| Muse Wonder | -0.8 |
| Claude Haiku 5.5 | -0.9 |
| Inkling | -1.0 |