FrontierMath Leaderboard
math60 models · Updated October 2026
On the FrontierMath benchmark, GPT-5.4 Pro ranks #1 with a score of 50.0, while GPT-5 Nano offers the best score-per-dollar at $0.40/1M output tokens. The full ranking, with cost per million tokens, is below.
💰 Best value
GPT-5 Nano — score 8.3 at $0.40/1M output tokens
| # | Model | Score | Output / 1M | Context |
|---|---|---|---|---|
| 1 | GPT-5.4 ProOpenAI | 50.0 | $180.00 | 1.1M |
| 2 | GPT-5.4OpenAI | 47.6 | $15.00 | 1.1M |
| 3 | Claude Opus 4.8Anthropic | 47.2 | $25.00 | 1M |
| 4 | Claude Opus 4.7Anthropic | 43.8 | $25.00 | 1M |
| 5 | Claude Opus 4.6Anthropic | 40.7 | $25.00 | 1M |
| 6 | GPT-5.2OpenAI | 40.7 | $14.00 | 400K |
| 7 | Muse SparkMeta AI | 39.0 | — | — |
| 8 | Gemini 3.5 FlashGoogle | 39.0 | $9.00 | 1M |
| 9 | Kimi K2.6MoonshotAI | 39.0 | $3.41 | 262K |
| 10 | Gemini 3 Pro PreviewGoogle | 37.6 | $12.00 | 1M |
| 11 | Gemini 3.1 Pro PreviewGoogle | 36.9 | $12.00 | 1M |
| 12 | Gemini 3 Flash PreviewGoogle | 35.6 | $3.00 | 1M |
| 13 | GLM 5.1Z.ai | 33.5 | $3.04 | 205K |
| 14 | GPT-5OpenAI | 32.4 | $10.00 | 400K |
| 15 | Claude Sonnet 4.6Anthropic | 32.4 | $15.00 | 1M |
| 16 | GPT-5.1OpenAI | 31.0 | $10.00 | 400K |
| 17 | GPT-5.4 MiniOpenAI | 28.3 | $4.50 | 400K |
| 18 | GPT-5 MiniOpenAI | 27.2 | $2.00 | 400K |
| 19 | Qwen3.6 PlusQwen | 26.2 | $1.95 | 1M |
| 20 | GPT-5.4 NanoOpenAI | 25.9 | $1.25 | 400K |
| 21 | o4 MiniOpenAI | 24.8 | $4.40 | 200K |
| 22 | Qwen3.6 Max PreviewAlibaba | 23.1 | $7.80 | 262K |
| 23 | Kimi K2 0711MoonshotAI | 21.4 | $2.30 | 131K |
| 24 | Qwen3.5 PlusAlibaba | 21.0 | $2.40 | 1M |
| 25 | Claude Opus 4.5Anthropic | 20.7 | $25.00 | 200K |
| 26 | Grok-4Poe | 19.7 | $15.00 | 256K |
| 27 | o3OpenAI | 18.7 | $8.00 | 200K |
| 28 | GLM 5Z.ai | 16.4 | $3.20 | 205K |
| 29 | Claude Sonnet 4.5Anthropic | 15.2 | $15.00 | 1M |
| 30 | Gemini 2.5 Pro (Jun 2025)Google DeepMind | 14.1 | — | — |
| 31 | o3 MiniOpenAI | 12.4 | $4.40 | 200K |
| 32 | Qwen3.6 FlashQwen | 10.3 | $1.13 | 1M |
| 33 | Gemini 2.5 Pro Preview 06-05Google | 10.3 | $10.00 | 1M |
| 34 | o1OpenAI | 9.3 | $60.00 | 200K |
| 35 | Qwen3 235B-A22BAlibaba | 8.5 | $2.80 | 131K |
| 36 | GPT-5 NanoOpenAI | 8.3 | $0.40 | 400K |
| 37 | Claude Opus 4.1Anthropic | 7.2 | $75.00 | 200K |
| 38 | Qwen3.5 FlashAlibaba | 6.2 | $0.40 | 1M |
| 39 | Claude Haiku 4.5Anthropic | 5.9 | $5.00 | 200K |
| 40 | GPT-4.1OpenAI | 5.5 | $8.00 | 1M |
| 41 | Gemini 2.5 FlashGoogle | 4.8 | $2.50 | 1M |
| 42 | Claude Opus 4Anthropic | 4.5 | $75.00 | 200K |
| 43 | GPT-4.1 MiniOpenAI | 4.5 | $1.60 | 1M |
| 44 | Claude Sonnet 4Anthropic | 4.1 | $15.00 | 200K |
| 45 | Claude 3.7 SonnetHelicone | 4.1 | $15.00 | 200K |
| 46 | GLM 4.6Z.ai | 3.8 | $2.00 | 205K |
| 47 | GLM 4.7Z.ai | 2.4 | $2.20 | 205K |
| 48 | Claude 3.5 SonnetDigitalOcean | 2.1 | $15.00 | 200K |
| 49 | Qwen-PlusQwen | 1.7 | $0.78 | 1M |
| 50 | DeepSeek V3DeepSeek | 1.7 | $0.28 | 164K |
| 51 | o1-miniAzure | 1.7 | $4.40 | 128K |
| 52 | GPT-4.1 NanoOpenAI | 1.0 | $0.40 | 1M |
| 53 | Qwen MaxAlibaba | 1.0 | $6.40 | 33K |
| 54 | Llama-4-Maverick-17B-128E-Instruct-FP8Vercel AI Gateway | 0.7 | $0.00 | 128K |
| 55 | Grok-2xAI | 0.7 | — | — |
| 56 | Mistral Medium (latest)Mistral | 0.3 | $7.50 | 262K |
| 57 | GPT-4oOpenAI | 0.3 | $10.00 | 128K |
| 58 | Claude 3.5 HaikuHelicone | 0.3 | $4.00 | 200K |
| 59 | Mistral LargeMistral AI | 0.3 | $6.00 | 128K |
| 60 | Llama 4 Scout 17B 16E InstructAzure | 0.0 | $0.78 | 128K |
Frequently asked questions
What is the FrontierMath benchmark?
FrontierMath is a math used to evaluate and rank large language models. This page lists 60 models by their FrontierMath score.
Which model is best on FrontierMath?
GPT-5.4 Pro (OpenAI) currently leads the FrontierMath leaderboard with a score of 50.0.
What is the best value model on FrontierMath?
GPT-5 Nano offers the best score-per-dollar on FrontierMath — a score of 8.3 at $0.40 per 1M output tokens.
Pricing is indicative — confirm with the provider before production use. Updated October 2026.