LMArena
Indexed results from LMArena. Each ranking retains its own metric and evaluation conditions.
LMArena Agent Bash Recovery Steps Overall
lmarena · LMArena · Arena score · score
- Anthropic: Claude Opus 5anthropic0.136
- Anthropic: Claude Opus 5anthropic0.126
- Anthropic: Claude Fable 5anthropic0.096
- OpenAI: GPT-5.5openai0.095
- Anthropic: Claude Sonnet 5anthropic0.074
- SpaceXAI: Grok 4.7x-ai0.072
- OpenAI: GPT-6 Solopenai0.07
- SpaceXAI: Grok 4.6x-ai0.057
- OpenAI: GPT-6 Astraopenai0.055
- OpenAI: GPT-6 Lunaopenai0.052
- SpaceXAI: Grok 4.5x-ai0.051
- OpenAI: GPT-6.1 Solopenai0.045
LMArena Agent Bash Reliability Overall
lmarena · LMArena · Arena score · score
- Anthropic: Claude Sonnet 5anthropic0.151
- OpenAI: GPT-6.1 Terraopenai0.134
- OpenAI: GPT-5.6 Solopenai0.114
- OpenAI: GPT-6 Solopenai0.109
- Anthropic: Claude Fable 5anthropic0.104
- SpaceXAI: Grok 4.6x-ai0.1
- Google: Gemini 3.8 Flashgoogle0.099
- Anthropic: Claude Opus 5anthropic0.098
- OpenAI: GPT-6 Lunaopenai0.095
- SpaceXAI: Grok 4.7x-ai0.093
- OpenAI: GPT-6 Astraopenai0.078
- OpenAI: GPT-6 Lunaopenai0.069
LMArena Agent Overall
lmarena · LMArena · Arena score · score
- Google: Gemini 4 Carbongoogle0.17
- OpenAI: GPT-6 Astraopenai0.145
- OpenAI: GPT-6.1 Solopenai0.125
- OpenAI: GPT-6.1 Terraopenai0.12
- Google: Gemini 4 Argongoogle0.107
- Anthropic: Claude Fable 5anthropic0.073
- Anthropic: Claude Opus 5anthropic0.072
- OpenAI: GPT-6 Solopenai0.069
- OpenAI: GPT-5.6 Solopenai0.046
- SpaceXAI: Grok 4.7x-ai0.026
- Google: Gemini 3.8 Flashgoogle0.023
- Anthropic: Claude Sonnet 5anthropic0.02
LMArena Agent Praise Complaint Overall
lmarena · LMArena · Arena score · score
- Google: Gemini 4 Carbongoogle0.41
- OpenAI: GPT-6 Astraopenai0.317
- OpenAI: GPT-6.1 Solopenai0.275
- Google: Gemini 4 Argongoogle0.241
- OpenAI: GPT-6.1 Terraopenai0.2
- Anthropic: Claude Fable 5anthropic0.143
- Anthropic: Claude Opus 5anthropic0.122
- OpenAI: GPT-6 Solopenai0.1
- OpenAI: GPT-5.6 Solopenai0.063
- Google: Gemini 3.8 Flashgoogle0.049
- SpaceXAI: Grok 4.7x-ai0.031
- Anthropic: Claude Sonnet 5anthropic-0.012
LMArena Agent Steerability Overall
lmarena · LMArena · Arena score · score
- Google: Gemini 4 Carbongoogle0.191
- OpenAI: GPT-6.1 Terraopenai0.166
- OpenAI: GPT-6 Astraopenai0.148
- OpenAI: GPT-6.1 Solopenai0.134
- Google: Gemini 4 Argongoogle0.132
- OpenAI: GPT-6 Solopenai0.101
- Anthropic: Claude Fable 5anthropic0.079
- OpenAI: GPT-5.6 Solopenai0.055
- Anthropic: Claude Opus 5anthropic0.044
- OpenAI: GPT-6 Lunaopenai0.035
- Anthropic: Claude Sonnet 5anthropic0.028
- OpenAI: GPT-6 Lunaopenai-0.001
LMArena Agent Task Outcome Explicit Overall
lmarena · LMArena · Arena score · score
- Google: Gemini 4 Carbongoogle0.184
- OpenAI: GPT-6 Astraopenai0.14
- Google: Gemini 4 Argongoogle0.132
- OpenAI: GPT-6.1 Solopenai0.117
- Anthropic: Claude Opus 5anthropic0.102
- OpenAI: GPT-6.1 Terraopenai0.097
- Anthropic: Claude Fable 5anthropic0.051
- SpaceXAI: Grok 4.7x-ai0.051
- OpenAI: GPT-6 Solopenai0.045
- OpenAI: GPT-5.6 Solopenai0.023
- Google: Gemini 3.8 Flashgoogle0.013
- Qwen: Qwen3.8 Flashqwen0.003
LMArena Agent Tool Hallucination Overall
lmarena · LMArena · Arena score · score
- OpenAI: GPT-6.1 Terraopenai0.01
- OpenAI: GPT-6 Astraopenai0.01
- SpaceXAI: Grok 4.6x-ai0.01
- OpenAI: GPT-6.1 Solopenai0.01
- OpenAI: GPT-6 Lunaopenai0.01
- OpenAI: GPT-6 Lunaopenai0.01
- OpenAI: GPT-5.6 Solopenai0.01
- OpenAI: GPT-6 Solopenai0.01
- SpaceXAI: Grok 4.7x-ai0.01
- OpenAI: GPT-6 Lunaopenai0.01
- OpenAI: GPT-5.6 Terraopenai0.01
- Anthropic: Claude Opus 5anthropic0.009
LMArena Document Overall
lmarena · LMArena · Arena rating · rating
- Anthropic: Claude Opus 5anthropic1,508.663
- Anthropic: Claude Fable 5anthropic1,496.608
- OpenAI: GPT-5.5openai1,484.293
- Google: Gemini 3.8 Flashgoogle1,481.924
- OpenAI: GPT-5.5openai1,480.896
- OpenAI: GPT-5.6 Solopenai1,480.037
- Google: Gemini 3.7 Flashgoogle1,476.472
- OpenAI: GPT-5.4openai1,472.946
- OpenAI: GPT-5.6 Terraopenai1,467.66
- OpenAI: GPT-6 Astraopenai1,467.252
- claude-opus-4-5@20251101anthropic1,465.714
- Google: Gemini 3.5 Flashgoogle1,461.768
LMArena Document Style Control Overall
lmarena · LMArena · Arena rating · rating
- Anthropic: Claude Fable 5anthropic1,500.119
- OpenAI: GPT-5.6 Solopenai1,486.489
- Anthropic: Claude Opus 5anthropic1,485.937
- OpenAI: GPT-5.5openai1,481.448
- OpenAI: GPT-5.5openai1,479.222
- OpenAI: GPT-5.6 Terraopenai1,474.173
- OpenAI: GPT-6 Astraopenai1,473.92
- OpenAI: GPT-5.4openai1,471.383
- Google: Gemini 3.8 Flashgoogle1,469.378
- Anthropic: Claude Sonnet 5anthropic1,468.879
- Google: Gemini 3.7 Flashgoogle1,467.94
- claude-opus-4-5@20251101anthropic1,467.496
LMArena Image Edit Agentic
lmarena · LMArena · Arena rating · rating
- OpenAI: GPT Image 2.5 Sunburstopenai1,540.094
- OpenAI: GPT Image 2.5 Flareopenai1,489.683
- OpenAI: GPT Image 2openai1,462.587
- xAI: Grok Imagine Image 2.0x-ai1,427.693
- Meta: Muse Imagemeta-llama1,387.193
- x-ai: Grok Imagine Imagex-ai1,314.589
- seedream-5.0-litebytedance1,288.939
- Alibaba: Wan 2.6 Imagealibaba1,194.783
LMArena Image Edit Multi Image Edit
lmarena · LMArena · Arena rating · rating
- OpenAI: GPT Image 2.5 Sunburstopenai1,524.731
- OpenAI: GPT Image 2.5 Flareopenai1,475.597
- OpenAI: GPT Image 2openai1,454.538
- OpenAI: GPT Image 2.6 Flareopenai1,439.06
- Google: Nano Banana 2.1google1,430.854
- Meta: Muse Imagemeta-llama1,398.687
- Gemini 3.1 Flash-Lite Image (Nano Banana 2 Lite)google1,289.366
- seedream-5.0-litebytedance1,266.197
- Alibaba: Wan 2.6 Imagealibaba1,220.26
- OpenAI: GPT Image 1openai1,126.377
- OpenAI: GPT Image 1 Miniopenai1,110.8
LMArena Image Edit Multi Image Edit Agentic
lmarena · LMArena · Arena rating · rating
- OpenAI: GPT Image 2.5 Sunburstopenai1,545.665
- OpenAI: GPT Image 2openai1,491.929
- OpenAI: GPT Image 2.5 Flareopenai1,480.606
- Meta: Muse Imagemeta-llama1,405.365
- seedream-5.0-litebytedance1,250.929
- Alibaba: Wan 2.6 Imagealibaba1,168.517
Models
LoadingBenchmarks
Explore every benchmark
LoadingThese charts contain model-linked results. The charts above also include unlinked source profiles. Compare scores within each evaluation. Model quality, usage, transparency and system performance measure different things.