Página automatizada. Redactada y traducida automáticamente por IA y verificada con las fuentes mediante software; sin editor humano. SI-IQ es un índice elaborado a partir de resultados públicos de benchmarks, no una prueba de CI. Cómo funciona
FrontierMath Tier 4: resultados de modelos de IA
FrontierMath Tier 4 es un benchmark de Matemáticas ejecutado por Epoch AI. En nuestros datos hay resultados publicados de 68 modelos.
Forma parte del índice SI-IQ (carga α = 2,32, nivel de azar 0 %).
- Evaluador
- Epoch AI
- Página principal del benchmark
- epoch.ai/frontiermath
Resultados
| 1 | GPT-6.1 Sol OpenAI | 100,0 % | 148144–152 | 9 |
| 2 | GPT-6 Astra OpenAI | 97,6 % | 150146–153 | 9 |
| 3 | Claude Opus 5.5 Anthropic | 95,0 % | 144141–148 | 9 |
| 4 | Claude Fable 5 Anthropic | 90,2 % | 138134–141 | 9 |
| 5 | GPT-6 Sol OpenAI | 90,0 % | 140137–144 | 9 |
| 6 | Claude Fable 5.1 Anthropic | 87,8 % | 140136–143 | 8 |
| 7 | GPT-5.6 Sol OpenAI | 82,9 % | 140137–143 | 9 |
| 8 | Claude Sonnet 5.5 Anthropic | 80,5 % | 140137–143 | 9 |
| 9 | GPT-5.5 Pro OpenAI | 78,0 % | 139135–142 | 5 datos insuficientes |
| 10 | Gemini 4 Argon Google DeepMind | 78,0 % | 140137–143 | 9 |
| 11 | AI Co-Mathematician Google DeepMind | 75,6 % | 137132–142 | 1 datos insuficientes |
| 12 | Claude Opus 5 Anthropic | 73,2 % | 136133–140 | 9 |
| 13 | GPT-5.5 OpenAI | 72,5 % | 138135–141 | 10 |
| 14 | GPT-5.6 Terra OpenAI | 70,7 % | 134131–137 | 9 |
| 15 | Muse Wonder Meta AI | 70,7 % | 136133–139 | 4 datos insuficientes |
| 16 | GPT-5.6 Luna OpenAI | 61,0 % | 130127–133 | 9 |
| 17 | GPT-5.4 Pro OpenAI | 58,5 % | 135131–138 | 5 datos insuficientes |
| 18 | Claude Opus 4.8 Anthropic | 56,1 % | 132129–135 | 9 |
| 19 | GPT-6 Luna OpenAI | 56,1 % | 130127–133 | 9 |
| 20 | GPT-5.4 OpenAI | 49,0 % | 131128–133 | 10 |
| 21 | Claude Haiku 5.5 Anthropic | 46,3 % | 129126–132 | 9 |
| 22 | Muse Spark 1.3 Meta AI | 46,3 % | 128125–131 | 5 datos insuficientes |
| 23 | Qwen 3.8 Max Alibaba | 46,3 % | 130127–133 | 7 datos insuficientes |
| 24 | GPT-5.2 Pro OpenAI | 46,0 % | 132128–135 | 3 datos insuficientes |
| 25 | Kimi K3abierto Moonshot | 39,0 % | 127124–130 | 9 |
| 26 | Gemini 3.7 Flash Google DeepMind | 36,6 % | 131128–134 | 9 |
| 27 | Qwen3.7-Max Alibaba | 34,1 % | 125122–128 | 10 |
| 28 | Qwen3.8 Max (0902) Alibaba | 34,1 % | 127123–130 | 6 datos insuficientes |
| 29 | Claude Opus 4.7 Anthropic | 31,7 % | 127124–130 | 10 |
| 30 | Grok 4.6 xAI | 31,7 % | 126123–129 | 9 |
| 31 | Muse Spark Openabierto Meta AI | 31,7 % | 127124–130 | 7 |
| 32 | GPT-5.2 OpenAI | 31,7 % | 125122–128 | 11 |
| 33 | Claude Sonnet 5 Anthropic | 29,3 % | 126123–129 | 9 |
| 34 | GLM-5.2abierto Z.ai (Zhipu AI) | 29,3 % | 121118–124 | 10 |
| 35 | GLM-5.3abierto Z.ai (Zhipu AI) | 29,3 % | 125122–128 | 9 |
| 36 | Claude Opus 4.6 Anthropic | 26,8 % | 123120–126 | 10 |
| 37 | DeepSeek V4 Pro 0813abierto DeepSeek | 26,8 % | 128125–131 | 9 |
| 38 | DeepSeek V4.1 Flashabierto DeepSeek | 26,8 % | 128125–131 | 9 |
| 39 | Gemini 3.1 Pro Google DeepMind | 26,8 % | 127124–130 | 10 |
| 40 | Gemini 3.5 Flash Google DeepMind | 26,8 % | 126123–129 | 10 |
| 41 | Mistral Large 4 Mistral AI | 26,8 % | 122119–125 | 7 datos insuficientes |
| 42 | Kimi K2.6abierto Moonshot | 25,6 % | 121118–124 | 10 |
| 43 | DeepSeek V4 Flash 0731abierto DeepSeek | 24,4 % | 124121–127 | 9 |
| 44 | Grok 4.5 xAI | 24,4 % | 124121–127 | 6 datos insuficientes |
| 45 | GPT-5 OpenAI | 22,0 % | 121118–124 | 12 |
| 46 | Gemini 3.6 Flash Google DeepMind | 22,0 % | 125122–128 | 9 |
| 47 | Gemini 3.8 Flash Google DeepMind | 22,0 % | 130127–133 | 9 |
| 48 | GPT-5 Pro OpenAI | 19,5 % | 124120–128 | 3 datos insuficientes |
| 49 | GLM-5.3-Flashabierto Z.ai (Zhipu AI) | 17,1 % | 120117–124 | 8 |
| 50 | Gemini 3 Flash Google DeepMind | 17,1 % | 122119–125 | 10 |
| 51 | Grok 4.7 xAI | 17,1 % | 123120–127 | 7 datos insuficientes |
| 52 | Inkling-Smallabierto Thinking Machines | 17,1 % | 118115–122 | 8 |
| 53 | Grok 4.3 Beta xAI | 14,6 % | 117114–121 | 6 datos insuficientes |
| 54 | GPT-5 mini OpenAI | 12,2 % | 115112–118 | 12 |
| 55 | GPT-5.4 Nano OpenAI | 12,2 % | 118115–121 | 9 |
| 56 | Kimi K2.7 Codeabierto Moonshot | 12,2 % | 118115–122 | 8 |
| 57 | GPT-5.4 Mini OpenAI | 9,8 % | 117113–120 | 9 |
| 58 | Claude Opus 4.5 Anthropic | 4,9 % | 115111–118 | 11 |
| 59 | Inklingabierto Thinking Machines | 4,9 % | 114110–117 | 8 |
| 60 | o4-mini OpenAI | 4,9 % | 112109–115 | 10 |
| 61 | Claude Opus 4.1 Anthropic | 2,4 % | 105102–109 | 6 datos insuficientes |
| 62 | Claude Sonnet 4.5 Anthropic | 2,4 % | 110107–113 | 12 |
| 63 | DeepSeek-V4-Proabierto DeepSeek | 2,4 % | 117113–120 | 10 |
| 64 | GPT-5 nano OpenAI | 2,4 % | 105102–109 | 11 |
| 65 | GPT-5.5 Instant OpenAI | 2,4 % | 107103–111 | 4 datos insuficientes |
| 66 | Gemini 2.5 Pro (Jun 2025) Google DeepMind | 0,0 % | 108104–112 | 8 |
| 67 | Gemini 3.5 Flash-Lite Google DeepMind | 0,0 % | 109106–113 | 8 |
| 68 | o3-mini OpenAI | 0,0 % | 105101–108 | 9 |
Fuentes de datos: Aider polyglot leaderboard data (Apache-2.0); Arena leaderboard dataset on Hugging Face (CC BY 4.0); Berkeley Function Calling Leaderboard, Gorilla project (Apache-2.0); Epoch AI, ‘Capabilities & benchmarking’ (CC BY 4.0); LiteLLM, model_prices_and_context_window.json (MIT); Source: OpenRouter (openrouter.ai), licensed under CC BY 4.0: reuse and republish with attribution to OpenRouter (licencias y cambios).