Benchmark · Arenas: people's votes
LMArena Agent Arena
How models do as agents in sessions people run through the arena, scored with inverse propensity scoring (IPS) from signals in those sessions (task outcome, bash reliability, steerability, tool hallucination, praise and complaints) rather than from pairwise votes.
- Score
- IPS score, higher is better
- Run by
- LMArena
- Status
- Current, newest result 18 November 2026
- Updated
- daily
- Data
- the published file, CC-BY-4.0, read 20 November 2026
Results
40 models we list, each at its best setting, best first.
| # | Model | Score | Range | Who ran it | Setting | Date | |
|---|---|---|---|---|---|---|---|
| 1 | Claude Fable 5.2 A | 0.274 | 0.245 to 0.304 | Independent | effort max | 2026-11-18 | source |
| 2 | Claude Opus 5.5 A | 0.186 | 0.166 to 0.207 | Independent | effort max | 2026-11-18 | source |
| 3 | Gemini 4 Neon B | 0.184 | 0.159 to 0.209 | Independent | effort high | 2026-11-18 | source |
| 4 | Gemini 4 Carbon B | 0.161 | 0.154 to 0.168 | Independent | effort high | 2026-11-18 | source |
| 5 | Claude Sonnet 5.5 A | 0.156 | 0.127 to 0.185 | Independent | effort max | 2026-11-18 | source |
| 6 | GPT-6.2 Astra AA | 0.152 | 0.125 to 0.180 | Independent | effort max | 2026-11-18 | source |
| 7 | GPT-6 Astra A | 0.146 | 0.125 to 0.166 | Independent | effort max | 2026-11-18 | source |
| 8 | Claude Fable 5.1 A | 0.137 | 0.116 to 0.157 | Independent | effort max | 2026-11-18 | source |
| 9 | Gemini 4 Hydrogen B | 0.135 | 0.118 to 0.153 | Independent | effort high | 2026-11-18 | source |
| 10 | GPT-6.1 Sol AA | 0.113 | 0.102 to 0.124 | Independent | effort max | 2026-11-18 | source |
| 10 | GPT-6.2 Sol AA | 0.113 | 0.089 to 0.137 | Independent | effort max | 2026-11-18 | source |
| 12 | GPT-6.1 Terra AA | 0.105 | 0.093 to 0.117 | Independent | effort max | 2026-11-18 | source |
| 13 | Gemini 4 Argon B | 0.103 | 0.094 to 0.112 | Independent | effort high | 2026-11-18 | source |
| 14 | GPT-6 Sol A | 0.065 | 0.049 to 0.080 | Independent | effort max | 2026-11-18 | source |
| 15 | Tencent Hy4 B | 0.061 | 0.049 to 0.073 | Independent | 2026-11-18 | source | |
| 16 | Kimi K3.1 Pro D | 0.039 | 0.028 to 0.050 | Independent | effort max | 2026-11-18 | source |
| 17 | Gemini 3.8 Flash AA | 0.015 | 0.006 to 0.024 | Independent | effort high | 2026-11-18 | source |
| 17 | Grok 4.7 BB | 0.015 | -0.000 to 0.030 | Independent | effort xhigh | 2026-11-18 | source |
| 17 | Inkling 2 Small C | 0.015 | 0.005 to 0.025 | Independent | 2026-11-18 | source | |
| 20 | Muse Spark 1.3 D | 0.011 | 0.002 to 0.020 | Independent | effort max | 2026-11-18 | source |
| 21 | Muse Ember D | 0.010 | 0.000 to 0.019 | Independent | effort max | 2026-11-18 | source |
| 22 | DeepSeek V4.1 Flash C | 0.008 | 0.003 to 0.013 | Independent | effort max | 2026-11-18 | source |
| 23 | GPT-6 Luna A | 0.006 | 0.002 to 0.011 | Independent | effort max | 2026-11-18 | source |
| 23 | Muse Wonder D | 0.006 | -0.007 to 0.019 | Independent | effort max | 2026-11-18 | source |
| 25 | GLM-5.3 B | -0.000 | -0.009 to 0.008 | Independent | effort max | 2026-11-18 | source |
| 26 | Kimi K3 B | -0.001 | -0.008 to 0.005 | Independent | effort max | 2026-11-18 | source |
| 27 | Claude Haiku 5.5 A | -0.006 | -0.016 to 0.004 | Independent | effort high | 2026-11-18 | source |
| 28 | MiniMax-M3.1-Flash C | -0.037 | -0.051 to -0.022 | Independent | 2026-11-18 | source | |
| 29 | GLM-5.3-Flash B | -0.039 | -0.046 to -0.031 | Independent | 2026-11-18 | source | |
| 30 | Qwen3.8-Flash-Next C | -0.049 | -0.057 to -0.041 | Independent | 2026-11-18 | source | |
| 31 | Step 5 Preview E | -0.051 | -0.066 to -0.037 | Independent | 2026-11-18 | source | |
| 32 | Qwen3.8-27B B | -0.055 | -0.064 to -0.046 | Independent | 2026-11-18 | source | |
| 33 | Gemini 3.1 Pro (preview) AA | -0.057 | -0.074 to -0.039 | Independent | 2026-11-18 | source | |
| 34 | Mistral Large 4 C | -0.071 | -0.092 to -0.051 | Independent | 2026-11-18 | source | |
| 35 | Hy3 B | -0.086 | -0.102 to -0.071 | Independent | 2026-11-18 | source | |
| 36 | MiniMax M3 C | -0.109 | -0.121 to -0.097 | Independent | 2026-11-18 | source | |
| 37 | Mistral Medium 3.5 B | -0.145 | -0.162 to -0.127 | Independent | 2026-11-18 | source | |
| 38 | Nemotron 3.5 Super VL C | -0.173 | -0.182 to -0.164 | Independent | precision NVFP4 | 2026-11-18 | source |
| 39 | Inkling B | -0.185 | -0.200 to -0.169 | Independent | 2026-11-18 | source | |
| 39 | Inkling-Small B | -0.185 | -0.206 to -0.165 | Independent | 2026-11-18 | source |
Every setting, 46 rows
| Model | As the leaderboard names it | Score | Who ran it | Setting | Date | |
|---|---|---|---|---|---|---|
| Claude Fable 5.2 | Claude Fable 5.2 (Max) | 0.274 | Independent | effort max | 2026-11-18 | source |
| Claude Opus 5.5 | Claude Opus 5.5 (Max) | 0.186 | Independent | effort max | 2026-11-18 | source |
| Gemini 4 Neon | Gemini 4 Neon (High) | 0.184 | Independent | effort high | 2026-11-18 | source |
| Gemini 4 Carbon | Gemini 4 Carbon (High) | 0.161 | Independent | effort high | 2026-11-18 | source |
| Claude Sonnet 5.5 | Claude Sonnet 5.5 (Max) | 0.156 | Independent | effort max | 2026-11-18 | source |
| GPT-6.2 Astra | GPT-6.2 Astra (Max) | 0.152 | Independent | effort max | 2026-11-18 | source |
| GPT-6 Astra | GPT-6 Astra (Max) | 0.146 | Independent | effort max | 2026-11-18 | source |
| Claude Fable 5.1 | Claude Fable 5.1 (Max) | 0.137 | Independent | effort max | 2026-11-18 | source |
| Claude Opus 5.5 | Claude Opus 5.5 (High) | 0.137 | Independent | effort high | 2026-11-18 | source |
| Gemini 4 Hydrogen | Gemini 4 Hydrogen (High) | 0.135 | Independent | effort high | 2026-11-18 | source |
| GPT-6.1 Sol | GPT-6.1 Sol (Max) | 0.113 | Independent | effort max | 2026-11-18 | source |
| GPT-6.2 Sol | GPT-6.2 Sol (Max) | 0.113 | Independent | effort max | 2026-11-18 | source |
| GPT-6.1 Terra | GPT-6.1 Terra (Max) | 0.105 | Independent | effort max | 2026-11-18 | source |
| Gemini 4 Argon | Gemini 4 Argon (High) | 0.103 | Independent | effort high | 2026-11-18 | source |
| GPT-6.1 Sol | GPT-6.1 Sol (Medium) | 0.082 | Independent | effort medium | 2026-11-18 | source |
| Claude Sonnet 5.5 | Claude Sonnet 5.5 (High) | 0.074 | Independent | effort high | 2026-11-18 | source |
| GPT-6 Sol | GPT-6 Sol (Max) | 0.065 | Independent | effort max | 2026-11-18 | source |
| Tencent Hy4 | Hy4 | 0.061 | Independent | 2026-11-18 | source | |
| GPT-6.1 Terra | GPT-6.1 Terra (Medium) | 0.046 | Independent | effort medium | 2026-11-18 | source |
| Kimi K3.1 Pro | Kimi K3.1 Pro (Max) | 0.039 | Independent | effort max | 2026-11-18 | source |
| Gemini 3.8 Flash | Gemini 3.8 Flash (High) | 0.015 | Independent | effort high | 2026-11-18 | source |
| Grok 4.7 | Grok 4.7 (xHigh) | 0.015 | Independent | effort xhigh | 2026-11-18 | source |
| Inkling 2 Small | Inkling 2 Small | 0.015 | Independent | 2026-11-18 | source | |
| Muse Spark 1.3 | Muse Spark 1.3 (Max) | 0.011 | Independent | effort max | 2026-11-18 | source |
| Muse Ember | Muse Ember (Max) | 0.010 | Independent | effort max | 2026-11-18 | source |
| DeepSeek V4.1 Flash | DeepSeek-V4.1-Flash-20261017 (Max) | 0.008 | Independent | effort max | 2026-11-18 | source |
| GPT-6 Luna | GPT-6 Luna (Max) | 0.006 | Independent | effort max | 2026-11-18 | source |
| Muse Wonder | Muse Wonder (Max) | 0.006 | Independent | effort max | 2026-11-18 | source |
| GLM-5.3 | GLM-5.3 (Max) | -0.000 | Independent | effort max | 2026-11-18 | source |
| Kimi K3 | Kimi K3 (Max) | -0.001 | Independent | effort max | 2026-11-18 | source |
| Claude Haiku 5.5 | Claude Haiku 5.5 (High) | -0.006 | Independent | effort high | 2026-11-18 | source |
| GPT-6 Luna | GPT-6 Luna (Medium) | -0.031 | Independent | effort medium | 2026-11-18 | source |
| MiniMax-M3.1-Flash | MiniMax-M3.1-Flash | -0.037 | Independent | 2026-11-18 | source | |
| GLM-5.3-Flash | GLM-5.3-Flash | -0.039 | Independent | 2026-11-18 | source | |
| Qwen3.8-Flash-Next | Qwen3.8-Flash-Next | -0.049 | Independent | 2026-11-18 | source | |
| Step 5 Preview | Step 5 Preview | -0.051 | Independent | 2026-11-18 | source | |
| Qwen3.8-27B | Qwen3.8-27B | -0.055 | Independent | 2026-11-18 | source | |
| Gemini 3.1 Pro (preview) | Gemini 3.1 Pro Preview | -0.057 | Independent | 2026-11-18 | source | |
| GPT-6 Luna | GPT-6 Luna (Low) | -0.064 | Independent | effort low | 2026-11-18 | source |
| Mistral Large 4 | Mistral Large 4 | -0.071 | Independent | 2026-11-18 | source | |
| Hy3 | Hy3 | -0.086 | Independent | 2026-11-18 | source | |
| MiniMax M3 | MiniMax-M3 | -0.109 | Independent | 2026-11-18 | source | |
| Mistral Medium 3.5 | Mistral Medium 3.5 | -0.145 | Independent | 2026-11-18 | source | |
| Nemotron 3.5 Super VL | NVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4) | -0.173 | Independent | precision NVFP4 | 2026-11-18 | source |
| Inkling | Inkling | -0.185 | Independent | 2026-11-18 | source | |
| Inkling-Small | Inkling-Small | -0.185 | Independent | 2026-11-18 | source |
A rank is among the models we list with a result, not a place on the whole leaderboard, and a model offered by several hosts is shown once, on its own listing. These results don't move our grades. Data: LMArena leaderboard dataset (lmarena-ai/leaderboard-dataset), CC BY 4.0, CC-BY-4.0.
For companies
Do agents find, use and choose your tools?
An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.