LMArena Agent Arena Code Net Improvement (Steering Burden)
Agents · 2026-08-14
Agent Arena Code category net-improvement aggregate: the overall board's five-signal net improvement recomputed on sessions tagged as writing or debugging code, automating workflows or analyzing data. A session can carry several category tags and every category session is also on the overall board, so this is a component of the overall Agent Arena family rather than a separate primary.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 23.4 |
| GPT-6 Astra | 16.1 |
| Claude Fable 5.1 | 15.1 |
| GPT-6.1 Sol | 14.7 |
| Claude Sonnet 5.5 | 14.6 |
| Muse Wonder Full | 11.9 |
| GPT-6 Sol | 11.2 |
| Gemini 4 Argon | 10.6 |
| Claude Fable 5 | 8.8 |
| Claude Opus 5 | 7.9 |
| DeepSeek V4.1 Flash 1017 | 7.2 |
| GPT-5.6 Sol | 6.9 |
| Sonnet 5 | 6.4 |
| GPT-5.5 | 5.2 |
| Claude Haiku 5.5 | 4.4 |
| GPT-6 Luna | 4.3 |
| Grok 4.7 | 4.1 |
| Muse Spark 1.3 | 4.1 |
| GLM-5.2 | 3.1 |
| DeepSeek V4.1 Flash | 3.1 |
| MiMo-V2.6-Pro | 2.8 |
| Hy4 Preview | 2.7 |
| Muse Wonder | 2.7 |
| GPT-5.4 | 2.5 |
| MiMo-V2.6-Flash | 2.4 |
| GLM-5.3 | 1.5 |
| Qwen3.8-Max | 1.4 |
| Kimi K3 | 1.4 |
| Grok 4.6 | 1.3 |
| Gemini 3.8 Flash | 1.1 |
| DeepSeek V4 Pro 0813 | 0.0 |
| GPT-5.6 Terra | 0.0 |
| Step 5 Preview | -0.7 |
| GPT-5.6 Luna | -1.7 |
| Qwen3.8-Flash-Next | -1.8 |
| GLM-5.3 Flash | -2.2 |
| Qwen3.8 27B | -3.1 |
| Gemini 3.7 Flash | -3.2 |
| Hy3 | -4.0 |
| Mistral Large 4 Preview | -4.1 |