AnotherWrapper
    $249
    LLMImageVideoSpeechTranscription

    MATH

    🏆 Leaderboard

    As of August 20, 2026, GLM-4.5 is #1 for MATH at 98.2%. Ranked by the MATH score 86 models in this index have a published MATH score. MATH leaderboard with live API prices. Tests mathematical problem solving across algebra, geometry, calculus, and number theory.

    Updated August 20, 2026282 models33 providers
    GLM-4.5OSS
    Z AI · Open Source
    		98.2%$2.80
    o3-mini
    OpenAI · Proprietary
    		97.9%$5.50
    GPT-5 mini
    OpenAI · Proprietary
    		97.8%$2.25
    o4-mini
    OpenAI · Proprietary
    		97.8%$5.50
    o3
    OpenAI · Proprietary
    		97.8%$10.00
    Claude Sonnet 4.5
    Anthropic · Proprietary
    		97.7%$18.00
    DeepSeek-R1OSS
    DeepSeek · Open Source
    		97.3%$2.74
    Gemini 2.5 Pro
    Google · Proprietary
    		97.1%$11.25
    DeepSeek-R1-0528OSS
    DeepSeek · Open Source
    		96.6%$2.74
    o1
    OpenAI · Proprietary
    		96.4%$75.00
    Claude Haiku 4.5
    Anthropic · Proprietary
    		96.4%$6.00
    Grok-3 Mini
    xAI · Proprietary
    		95.8%$0.80
    GPT-5 nano
    OpenAI · Proprietary
    		95.2%$0.45
    DeepSeek-V3OSS
    DeepSeek · Open Source
    		94%$1.37
    Claude 3.7 Sonnet
    Anthropic · Proprietary
    		91.2%$18.00
    Mistral Large 3OSS
    Mistral · Open Source · via Mistral AI
    		90.4%$7.00
    MiniStral 3 (14B Instruct 2512)OSS
    Mistral · Open Source · via OpenRouter
    		90.4%$0.40
    o1-mini
    OpenAI · Proprietary
    		90%$15.00
    DeepSeek R1 Distill Llama 70BOSS
    DeepSeek · Open Source
    		89.9%$0.50
    Gemini 2.0 Flash
    Google · Proprietary
    		89.7%$0.50
    Kimi K2 0905
    Moonshot AI · Proprietary
    		89.1%$3.10
    Gemma 3 27BOSS
    Google · Open Source
    		89%$0.30
    Ministral 3 (8B Instruct 2512)OSS
    Mistral · Open Source · via OpenRouter
    		87.6%$0.30
    GPT-4.1 mini
    OpenAI · Proprietary
    		87.3%$2.00
    Jamba 1.5 LargeOSS
    AI21 Labs · Open Source
    		87%$10.00
    Gemini 2.0 Flash-Lite
    Google · Proprietary
    		86.8%$0.37
    Gemini 1.5 Pro
    Google · Proprietary
    		86.5%$12.50
    MiMo-V2.5-ProOSS
    Xiaomi · Open Source
    		86.2%$1.30
    o1-preview
    OpenAI · Proprietary
    		85.5%$75.00
    Claude Opus 4
    Anthropic · Proprietary
    		85.0%$90.00
    GLM-4.7OSS
    Z AI · Open Source
    		84.9%$2.80
    GPT-5
    OpenAI · Proprietary
    		84.7%$11.25
    Claude Sonnet 4
    Anthropic · Proprietary
    		84.4%$18.00
    Gemma 3 12BOSS
    Google · Open Source
    		83.8%$0.15
    Qwen2.5 72B InstructOSS
    Qwen · Open Source
    		83.1%$0.75
    GPT-4.1
    OpenAI · Proprietary
    		83.0%$10.00
    Ministral 3 (3B Instruct 2512)OSS
    Mistral · Open Source · via OpenRouter
    		83%$0.20
    Phi 4OSS
    Microsoft · Open Source
    		80.4%$0.21
    Claude 3.5 Sonnet
    Anthropic · Proprietary
    		78.3%$18.00
    Gemini 1.5 Flash
    Google · Proprietary
    		77.9%$0.75
    Llama 3.3 70B InstructOSS
    Meta · Open Source
    		77%$0.40
    Nova Pro
    Amazon · Proprietary
    		76.6%$4.00
    GPT-4o
    OpenAI · Proprietary
    		76.6%$12.50
    Grok-2
    xAI · Proprietary
    		76.1%$12.00
    Jamba 1.5 MiniOSS
    AI21 Labs · Open Source
    		75.8%$0.60
    Gemma 3 4BOSS
    Google · Open Source
    		75.6%$0.06
    DeepSeek-V3 0324OSS
    DeepSeek · Open Source
    		75.5%$1.42
    Qwen2.5 7B InstructOSS
    Qwen · Open Source
    		75.5%$0.60
    DeepSeek-V2.5OSS
    DeepSeek · Open Source
    		74.7%$0.42
    Llama 3.1 405B InstructOSS
    Meta · Open Source
    		73.8%$1.78
    Showing 1–50 of 282 models

    Next step

    You found the model. Now ship the product.

    Auth, billing, and the API layer are already decided. Fork 8 finished AI apps, or have us build the first version with you.

    Get the stack — $249Have us build it with you

    The index

    All Large Language Models

    282 models across 33 providers. Search or jump to a lab — every model page stays linked here.

    OpenAI51Anthropic22Google25Cursor2xAI15DeepSeek16MiniMax6Qwen38AI21 Labs2Amazon5Baidu1ByteDance2Cohere3IBM1Inception1inclusionAI1LG AI Research1Meituan4Meta12Microsoft5Mistral23Moonshot AI9Nous Research1NVIDIA5Poolside2Sakana AI1StepFun1Tencent1Thinking Machines1Unisound1Upstage1Xiaomi3Z AI20

    Baidu

    1 models
    ERNIE 4.5

    Inception

    1 models
    Mercury 2

    inclusionAI

    1 models
    Ling-3.0-flash

    LG AI Research

    1 models
    K-EXAONE-236B-A23B

    Nous Research

    1 models
    Hermes 3 70B

    Sakana AI

    1 models
    Sakana Namazu

    StepFun

    1 models
    Step-3.5-Flash

    Tencent

    1 models
    Hy3

    Thinking Machines

    1 models
    Inkling-Small

    Unisound

    1 models
    U2

    Upstage

    1 models
    Solar Pro 4

    Which model leads MATH right now?

    As of August 20, 2026, GLM-4.5 by Z AI is #1 for MATH at 98.2%. Ranked by the MATH score This board also tracks MATH. Next on the same board: o3-mini and GPT-5 mini. This math leaderboard ranks models by MATH. Scores come from public evals. Prices are the live API rates in the table above.

    Sources: OpenAI API pricing (https://developers.openai.com/api/docs/pricing); Anthropic Claude API pricing (https://platform.claude.com/docs/en/about-claude/pricing); Gemini API pricing (https://ai.google.dev/gemini-api/docs/pricing)

    Top 8 for math. Ranked by the MATH score Input and output are dollars per million tokens.
    RankModelMATHInput /MOutput /M
    1GLM-4.598.2%$0.60$2.20
    2o3-mini97.9%$1.10$4.40
    3GPT-5 mini97.8%$0.25$2.00
    4o4-mini97.8%$1.10$4.40
    5o397.8%$2.00$8.00
    6Claude Sonnet 4.597.7%$3.00$15.00
    7DeepSeek-R197.3%$0.55$2.19
    8Gemini 2.5 Pro97.1%$1.25$10.00

    AIME vs FrontierMath

    AIME is contest math. MATH-500 is textbook problems. FrontierMath is the set that is supposed to stay difficult. If every model is in the 90s on AIME, sort FrontierMath. That is the gap people actually mean by best AI for math in 2026.

    Related LLM leaderboards

    LLM LeaderboardCodingAgenticReasoningWritingScienceFinanceLong ContextVisionOpen Source

    Individual benchmark pages

    Rank one eval at a time. All LLM benchmarks.

    HLESWE BenchHumanity's Last Exam (No Tools)Humanity's Last Exam (With Tools)Terminal-Bench 2.0Terminal-BenchTerminal-Bench 2.1Terminal-Bench 2Terminal-Bench 3.0ARC-AGI-1 Verified

    MATH FAQ

    Who ranks #1 on the MATH leaderboard?

    As of August 20, 2026, GLM-4.5 by Z AI ranks #1 on MATH at 98.2%. API pricing is $0.60/M input and $2.20/M output.

    What are the top models on MATH?

    The current MATH ranking as of August 20, 2026 is 1. GLM-4.5 at 98.2%; 2. o3-mini at 97.9%; 3. GPT-5 mini at 97.8%.

    Which math model is the cheapest?

    Llama 3.2 3B Instruct is the cheapest scored model on this math leaderboard at $0.01/M input and $0.02/M output ($0.03 blended). GLM-4.5 still leads MATH at 98.2%.

    Should I always pick the #1 MATH model?

    Not automatically. GLM-4.5 leads MATH, but a cheaper scored model can be the better production choice if the quality gap is small. Use the table to weigh MATH against input/output price, context window, and related evals.

    How often is the MATH leaderboard updated?

    Scores and API prices on this page are refreshed from published evals and provider rates. The snapshot is labeled August 20, 2026. Treat it as a current index, not a one-off blog post.

    What is MATH?

    Tests mathematical problem solving across algebra, geometry, calculus, and number theory. This page ranks models that have published a MATH score, with live API token prices on the same row.

    Where is the MATH leaderboard?

    This page is the MATH leaderboard. Models are sorted by MATH, with input and output token prices on the same row so you can weigh score against cost. Official boards often omit price; that comparison is the point of this index.

    How is this MATH ranking different from the official board?

    Official eval pages own the methodology. This page keeps the published MATH score next to live API $/M so you can pick a production SKU, not only a trophy number.

    What is FrontierMath?

    FrontierMath is a research-level math eval built to stay hard after models saturate AIME and MATH-500. Use it for the extreme tail, not for high-school contest bragging.

    What does AIME 2025 measure?

    AIME 2025 is American Invitational Mathematics Examination problems. It is the common public contest-math signal. Many frontier models now score very high, which is why FrontierMath exists.

    Is the best AI for homework the same as the best math model?

    People search best AI for homework. This page ranks contest and research math, not a cheat sheet. A high AIME score means contest reasoning. It does not mean the model should do someone's take-home.

    ChatGPT vs Claude for math: which is better?

    Rank AIME, MATH-500, and FrontierMath on this table. The lead moves. Token price still matters if you are sampling many solutions.