Arena Elo — Math

The Arena text leaderboard's Math category: Elo ratings computed only from anonymous votes on maths-related prompts.

Also known as: Chatbot Arena Math leaderboard, LMArena Math category

unassessed

This page is a discovery lead. Nobody has yet assessed it against the catalogue contract, so it carries no disposition. Absence of evidence here is not evidence of staleness.
Categoryhuman-preference
Subcategorypairwise human preference, math prompts
Page statusactive
MetricElo
Directionhigher_is_better
PublisherLMSYS (Large Model Systems Organization), UC Berkeley Sky Computing Lab (founding org); operates today as Arena (formerly LMArena)

What it measures

arena_elo_math restricts the same live, anonymous pairwise-vote pool behind the Arena text leaderboard to conversations tagged as maths prompts, then fits the platform's usual Bradley-Terry rating on that subset alone. It is a category filter selected from the leaderboard's own Category view over the same underlying vote pool as arena_elo_overall, not a separately collected benchmark and not a graded-answer test: voters compare two chat responses to a maths question and pick the one they prefer, without the platform verifying which answer is numerically correct.

Task format

Anonymous, randomized side-by-side text chat restricted to prompts tagged as maths; a user votes for the preferred response.

Models reporting this benchmark

These figures come from the model cards, which carry one collection date per card and no per-score attribution. They are shown as reported, not as verified evidence.
ModelProviderScoreCard as of
GLM 5Zhipu AI1458.3undated
GLM 5.1Z.ai (Zhipu AI)1457.42026-04
o3-proOpenAI14402026-04
o3OpenAI1430.02026-04
o3-deep-researchOpenAI14302026-04
GLM 4.7Zhipu AI1429.7undated
GLM 4.6Zhipu AI1421.6undated
GLM 4.5Zhipu AI1415.0undated
DeepSeek R1 0528DeepSeek1410.02026-04
DeepSeek R1 0528 NVFP4 v2NVIDIA1410.02026-04
o4-miniOpenAI1410.02026-04
o4-mini-deep-researchOpenAI1410.02026-04
Claude Opus 4Anthropic1400.02026-04
Claude Opus 4.6Anthropic1400.02026-04
DeepSeek R1DeepSeek1400.02026-04
DeepSeek R1 Distill Llama 70BDeepSeek1400.02026-04
DeepSeek R1 Distill Llama 8BDeepSeek14002026-04
DeepSeek R1 Distill Qwen 14BDeepSeek1400.02026-04
DeepSeek R1 Distill Qwen 32BDeepSeek1400.02026-04
DeepSeek R1 Distill Qwen 7BDeepSeek14002026-04
DeepSeek ReasonerDeepSeek1400.02026-04
Gemini 2.5 ProGoogle DeepMind1400.02026-04
Gemini 2.5 Pro Preview 05-06Google DeepMind1400.02026-04
Gemini 2.5 Pro Preview 06-05Google DeepMind1400.02026-04
Gemini 2.5 Pro Preview TTSGoogle DeepMind1400.02026-04
o1-proOpenAI1400.02026-04
Claude Opus 4.5Anthropic1390.02026-04
Claude Opus 4.5 (latest)Anthropic1390.02026-04
GLM 4.5 AirZhipu AI1390.0undated
GPT-5.1OpenAI1390.02026-04
GPT-5.1 ChatOpenAI1390.02026-04
GPT-5.1 CodexOpenAI1390.02026-04
GPT-5.1 Codex MaxOpenAI1390.02026-04
GPT-5.1 Codex miniOpenAI1390.02026-04
Claude Opus 4.1Anthropic1380.02026-04
Claude Opus 4.1 (latest)Anthropic1380.02026-04
GPT-5OpenAI1380.02026-04
GPT-5 Chat (latest)OpenAI1380.02026-04
GPT-5 ProOpenAI1380.02026-04
GPT-5-CodexOpenAI1380.02026-04
GPT-5.2OpenAI1380.02026-04
GPT-5.2 ChatOpenAI1380.02026-04
GPT-5.2 CodexOpenAI1380.02026-04
GPT-5.2 ProOpenAI1380.02026-04
GPT-5.3 Chat (latest)OpenAI1380.02026-04
GPT-5.3 CodexOpenAI1380.02026-04
GPT-5.3 Codex SparkOpenAI1380.02026-04
GPT-5.4OpenAI1380.02026-04
GPT-5.4 miniOpenAI1380.02026-04
GPT-5.4 nanoOpenAI1380.02026-04
GPT-5.4 ProOpenAI1380.02026-04
Grok 4xAI1380.02026-04
Grok 4 FastxAI1380.02026-04
Grok 4 Fast (Non-Reasoning)xAI1380.02026-04
Grok 4.1 FastxAI1380.02026-04
Grok 4.1 Fast (Non-Reasoning)xAI1380.02026-04
Grok 4.20 (Non-Reasoning)xAI1380.02026-04
Grok 4.20 (Reasoning)xAI1380.02026-04
Grok 4.20 Multi-AgentxAI1380.02026-04
Claude Sonnet 4Anthropic1370.02026-04
Claude Sonnet 4.6Anthropic1370.02026-04
o3-miniOpenAI1370.02026-04
GLM 4.7 FlashZhipu AI1367.8undated
Claude Opus 4 (latest)Anthropic1360.02026-04
Claude Sonnet 4.5Anthropic1360.02026-04
Claude Sonnet 4.5 (latest)Anthropic1360.02026-04
o1OpenAI1360.02026-04
o1-previewOpenAI1360.02026-04
Qwen3 235B-A22BAlibaba / Qwen Team1360.02026-04
GLM 4.5VZhipu AI1357.7undated
Gemini 2.5 FlashGoogle DeepMind1350.02026-04
Gemini 2.5 Flash ImageGoogle DeepMind1350.02026-04
Gemini 2.5 Flash Image (Preview)Google DeepMind1350.02026-04
Gemini 2.5 Flash LiteGoogle DeepMind1350.02026-04
Gemini 2.5 Flash Lite Preview 06-17Google DeepMind1350.02026-04
Gemini 2.5 Flash Lite Preview 09-25Google DeepMind1350.02026-04
Gemini 2.5 Flash Preview 04-17Google DeepMind1350.02026-04
Gemini 2.5 Flash Preview 05-20Google DeepMind1350.02026-04
Gemini 2.5 Flash Preview 09-25Google DeepMind1350.02026-04
Gemini 2.5 Flash Preview TTSGoogle DeepMind1350.02026-04
Qwen3-Coder 480B-A35B InstructAlibaba / Qwen Team1340.02026-04
Claude Sonnet 4 (latest)Anthropic1330.02026-04
DeepSeek V3.2DeepSeek1330.02026-04
DeepSeek V3.2 ExpDeepSeek1330.02026-04
GPT-5 MiniOpenAI1330.02026-04
o1-miniOpenAI1330.02026-04
Qwen3 32BAlibaba / Qwen Team1330.02026-04
Qwen3 32B AWQAlibaba / Qwen Team1330.02026-04
Qwen3 32B NVFP4NVIDIA1330.02026-04
DeepSeek V3.1DeepSeek1320.02026-04
GPT-4.1OpenAI1320.02026-04
Magistral Medium (latest)Mistral AI1320.02026-04
Claude Sonnet 3.7Anthropic1310.02026-04
DeepSeek V3DeepSeek1310.02026-04
DeepSeek V3 0324DeepSeek1310.02026-04
Grok 3xAI1310.02026-04
Grok 3 FastxAI1310.02026-04
Grok 3 Fast LatestxAI1310.02026-04
Grok 3 LatestxAI1310.02026-04
Claude Haiku 4.5Anthropic1300.02026-04
Claude Haiku 4.5 (latest)Anthropic1300.02026-04
DeepSeek ChatDeepSeek1300.02026-04
DeepSeek V2DeepSeek13002026-04
DeepSeek V2 LiteDeepSeek13002026-04
DeepSeek V2 Lite ChatDeepSeek13002026-04
Grok 3 MinixAI1300.02026-04
Grok 3 Mini FastxAI1300.02026-04
Grok 3 Mini Fast LatestxAI1300.02026-04
Grok 3 Mini LatestxAI1300.02026-04
Llama 4 Maverick 17B 128E InstructMeta1300.02026-04
Llama-4-Maverick-17B-128E-Instruct-FP8Meta1300.02026-04
Qwen3 30B A3B Instruct 2507Alibaba / Qwen Team1300.02026-04
Qwen3 30B A3B NVFP4NVIDIA1300.02026-04
Qwen3 30B-A3BAlibaba / Qwen Team1300.02026-04
Claude Sonnet 3.5Anthropic1290.02026-04
Claude Sonnet 3.5 v2Anthropic1290.02026-04
Gemini 2.0 FlashGoogle DeepMind1290.02026-04
GPT-4.1 miniOpenAI1290.02026-04
Qwen3 14BAlibaba / Qwen Team12902026-04
Qwen3 14B AWQAlibaba / Qwen Team12902026-04
Qwen3 14B NVFP4NVIDIA12902026-04
Gemma 4 31BGoogle DeepMind1280.02026-04
gemma 4 31B itGoogle DeepMind1280.02026-04
gemma 4 31B it GGUFUnsloth1280.02026-04
Gemma 4 31B IT NVFP4NVIDIA1280.02026-04
GPT-5 NanoOpenAI1280.02026-04
Magistral SmallMistral AI1280.02026-04
Magistral Small 2506Mistral AI1280.02026-04
Gemma 4 26BGoogle DeepMind1270.02026-04
gemma 4 26B A4B itGoogle DeepMind1270.02026-04
gemma 4 26B A4B it GGUFUnsloth1270.02026-04
GPT-4oOpenAI1270.02026-04
GPT-4o (2024-05-13)OpenAI1270.02026-04
GPT-4o (2024-08-06)OpenAI1270.02026-04
GPT-4o (2024-11-20)OpenAI1270.02026-04
Llama 4 Scout 17B 16EMeta1270.02026-04
Llama 4 Scout 17B 16E InstructMeta1270.02026-04
Llama-4-Scout-17B-16E-Instruct-FP8Meta1270.02026-04
Qwen2.5 72B InstructAlibaba / Qwen Team1270.02026-04
Mistral Large (latest)Mistral AI1260.02026-04
Mistral Large 2.1Mistral AI1260.02026-04
Mistral Large 3Mistral AI1260.02026-04
phi 4Microsoft1260.02026-04
Phi 4 multimodal instructMicrosoft12602026-04
Qwen3 8BAlibaba / Qwen Team12602026-04
Qwen3 8B AWQAlibaba / Qwen Team12602026-04
Qwen3 8B BaseAlibaba / Qwen Team12602026-04
Claude Haiku 3.5Anthropic1250.02026-04
Claude Haiku 3.5 (latest)Anthropic1250.02026-04
Command ACohere12502026-04
Command A ReasoningCohere12502026-04
Gemini 2.0 Flash LiteGoogle DeepMind1250.02026-04
Gemma 3 27BGoogle DeepMind1250.02026-04
GPT-4 TurboOpenAI1250.02026-04
Grok 2xAI12502026-04
Grok 2 (1212)xAI12502026-04
Grok 2 LatestxAI12502026-04
Grok 2 VisionxAI12502026-04
Grok 2 Vision (1212)xAI12502026-04
Grok 2 Vision LatestxAI12502026-04
Llama 3.1 405BMeta1250.02026-04
Llama 3.1 405B FP8Meta1250.02026-04
Llama 3.1 405B InstructMeta1250.02026-04
Llama 3.1 405B Instruct FP8Meta1250.02026-04
Claude Opus 3Anthropic1240.02026-04
Codestral (latest)Mistral AI1240.02026-04
Gemini 1.5 ProGoogle DeepMind1240.02026-04
GPT-4.1 nanoOpenAI1240.02026-04
GPT-4o miniOpenAI1240.02026-04
Llama 3.3 70B Instruct NVFP4NVIDIA1240.02026-04
Llama-3.3-70B-InstructMeta1240.02026-04
Qwen2.5 32B InstructAlibaba / Qwen Team12402026-04
Qwen2.5 32B Instruct AWQAlibaba / Qwen Team12402026-04
Mistral Medium (latest)Mistral AI1230.02026-04
Mistral Medium 3Mistral AI1230.02026-04
Mistral Medium 3.1Mistral AI1230.02026-04
Gemma 3 12BGoogle DeepMind1220.02026-04
GPT-4OpenAI1220.02026-04
Llama 3.1 70BMeta1220.02026-04
Llama 3.1 70B InstructMeta1220.02026-04
Phi 4 mini instructMicrosoft1220.02026-04
Qwen3 4BAlibaba / Qwen Team12202026-04
Qwen3 4B BaseAlibaba / Qwen Team12202026-04
Qwen3 4B Instruct 2507Alibaba / Qwen Team12202026-04
Qwen3 4B Instruct 2507 FP8Alibaba / Qwen Team12202026-04
Command R+Cohere1210.02026-04
Gemini 1.5 FlashGoogle DeepMind1210.02026-04
Gemini 1.5 Flash-8BGoogle DeepMind1210.02026-04
gemma 2 27B itGoogle DeepMind1210.02026-04
Mistral Small (latest)Mistral AI1200.02026-04
Mistral Small 24B Instruct 2501Mistral AI1200.02026-04
Mistral Small 3.1 24B Instruct 2503Mistral AI1200.02026-04
Mistral Small 3.2Mistral AI1200.02026-04
Mistral Small 3.2 24B Instruct 2506Mistral AI1200.02026-04
Mistral Small 4Mistral AI1200.02026-04
Mistral Small 4 119B 2603Mistral AI1200.02026-04
Claude Sonnet 3Anthropic1180.02026-04
Command RCohere11802026-04
gemma 2 9BGoogle DeepMind1180.02026-04
gemma 2 9B itGoogle DeepMind1180.02026-04

Showing the top 200 of 209.

Data

This page as JSON · Edit on GitHub