LMArena Agent Arena Work Tool Hallucination
Agents · 2026-08-14
Signed percentage-point Agent Arena Work diagnostic improvement signal for reductions in apparent nonexistent or inappropriate tool use, recomputed on sessions tagged as office work such as document creation, professional research, planning and professional writing. 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 | 0.7 |
| Kimi K3 | 0.7 |
| GLM-5.2 | 0.7 |
| GPT-6 Astra | 0.7 |
| Claude Opus 5 | 0.7 |
| GLM-5.3 Flash | 0.7 |
| GLM-5.3 | 0.7 |
| Grok 4.7 | 0.6 |
| Claude Fable 5 | 0.6 |
| Step 5 Preview | 0.6 |
| Claude Fable 5.1 | 0.6 |
| Muse Spark 1.3 | 0.6 |
| GPT-6 Sol | 0.6 |
| GPT-5.6 Sol | 0.6 |
| Grok 4.6 | 0.6 |
| DeepSeek V4 Pro 0813 | 0.6 |
| GPT-5.5 | 0.6 |
| GPT-6 Luna | 0.6 |
| Claude Sonnet 5.5 | 0.6 |
| Gemini 3.8 Flash | 0.6 |
| Gemini 3.7 Flash | 0.5 |
| DeepSeek V4.1 Flash | 0.5 |
| GPT-5.6 Terra | 0.5 |
| GPT-5.6 Luna | 0.5 |
| Claude Opus 5.5 | 0.5 |
| DeepSeek V4.1 Flash 1017 | 0.5 |
| GPT-5.4 | 0.4 |
| Sonnet 5 | 0.4 |
| Muse Wonder Full | 0.4 |
| Muse Wonder | 0.3 |
| Gemini 4 Argon | 0.3 |
| Muse Spark 1.1 | 0.3 |
| Muse Spark 1.2 | 0.2 |
| Hy4 Preview | 0.1 |
| Qwen3.8 27B | 0.1 |
| Gemini 3.1 Pro Preview | -0.0 |
| MiMo-V2.6-Pro | -0.1 |
| Qwen3.7-Max (source unspecified) | -0.2 |
| Inkling | -0.7 |
| Qwen3.8-Flash-Next | -0.8 |