Best model for your job # Best AI model for hard reasoning > For reasoning, GPT-6 Astra scores highest on the current data (87.6), weighting reasoning 70%, math 30%. The cheapest model in the top 10 is GPT-6.1 Sol at $2 - Canonical page: https://noometry.com/best/for/reasoning - Last updated: 2026-10-10 - Title: Best AI model for hard reasoning (October 2026) | Noometry For reasoning, GPT-6 Astra scores highest on the current data (87.6), weighting reasoning 70%, math 30%. The cheapest model in the top 10 is GPT-6.1 Sol at $2 / $10 per million tokens. Last verified October 10, 2026 Weights: [Reasoning](https://noometry.com/best/reasoning) 70%, [Math](https://noometry.com/best/math) 30%. Novel problem solving is measured best by abstraction puzzles and math. Best AI model for hard reasoning | # | Model | Provider | Fit | Reasoning | Math | Input $/M | Output $/M | | --- | --- | --- | --- | --- | --- | --- | --- | | 1 | [GPT-6 Astra](https://noometry.com/models/gpt-6-astra) | OpenAI | **87.6** | 85.1 | 93.5 | $10 | $50 | | 2 | [GPT-6.1 Sol](https://noometry.com/models/gpt-6-1-sol) | OpenAI | **85.4** | 81.9 | 93.7 | $2 | $10 | | 3 | [Claude Opus 5.5](https://noometry.com/models/claude-opus-5-5) | Anthropic | **83.7** | 80.2 | 91.8 | $4 | $20 | | 4 | [Claude Fable 5.1](https://noometry.com/models/claude-fable-5-1) | Anthropic | **80.6** | 76.7 | 89.6 | $10 | $50 | | 5 | [Claude Fable 5](https://noometry.com/models/claude-fable-5) | Anthropic | **80.3** | 76.8 | 88.5 | $10 | $50 | | 6 | [Claude Opus 5](https://noometry.com/models/claude-opus-5) | Anthropic | **79.9** | 77.2 | 86.2 | $5 | $25 | | 7 | [GPT-5.6 Sol](https://noometry.com/models/gpt-5-6-sol) | OpenAI | **78.0** | 74.8 | 85.6 | $4 | $20 | | 8 | [GPT-6 Sol](https://noometry.com/models/gpt-6-sol) | OpenAI | **78.0** | 74.0 | 87.2 | $2 | $10 | | 9 | [GPT-5.5 Pro](https://noometry.com/models/gpt-5-5-pro) | OpenAI | **76.5** | 73.3 | 84.0 | $30 | $180 | | 10 | [GPT-5.5](https://noometry.com/models/gpt-5-5) | OpenAI | **75.4** | 72.8 | 81.7 | $5 | $30 | | 11 | [Gemini 3.8 Flash](https://noometry.com/models/gemini-3-8-flash) | Google | **73.4** | 76.9 | 65.3 | $0.75 | $3.75 | | 12 | [GPT-5.4 Pro](https://noometry.com/models/gpt-5-4-pro) | OpenAI | **71.2** | 70.7 | 72.4 | $30 | $180 | | 13 | [Gemini 3.7 Flash](https://noometry.com/models/gemini-3-7-flash) | Google | **69.9** | 70.0 | 69.6 | $0.75 | $3.75 | | 14 | [Gemini 3.1 Pro Preview](https://noometry.com/models/gemini-3-1-pro-preview) | Google | **68.8** | 71.7 | 62.1 | $2 | $12 | | 15 | [Claude Opus 4.8](https://noometry.com/models/claude-opus-4-8) | Anthropic | **68.8** | 64.7 | 78.4 | $5 | $25 | | 16 | [GPT-5.6 Terra](https://noometry.com/models/gpt-5-6-terra) | OpenAI | **66.9** | 60.7 | 81.6 | $2 | $12 | | 17 | [Kimi K3](https://noometry.com/models/kimi-k3) (open weights) | Moonshot AI | **66.4** | 63.0 | 74.2 | $3 | $15 | | 18 | [GPT-5.4](https://noometry.com/models/gpt-5-4) | OpenAI | **65.3** | 61.8 | 73.5 | $2.50 | $15 | | 19 | [Claude Sonnet 5.5](https://noometry.com/models/claude-sonnet-5-5) | Anthropic | **64.2** | 54.0 | 87.9 | $2 | $10 | | 20 | [Grok 4.6](https://noometry.com/models/grok-4-6) | xAI | **63.1** | 61.4 | 67.0 | $2 | $6 | | 21 | [Gemini 3.5 Flash](https://noometry.com/models/gemini-3-5-flash) | Google | **62.2** | 62.8 | 60.7 | $1.50 | $9 | | 22 | [Qwen3.8 Max](https://noometry.com/models/qwen3-8-max) | Alibaba (Qwen) | **60.0** | 54.4 | 73.2 | $2 | $6 | | 23 | [Muse Spark 1.3](https://noometry.com/models/muse-spark-1-3) | Meta | **59.8** | 54.0 | 73.1 | $1.25 | $4.25 | | 24 | [Claude Opus 4.6](https://noometry.com/models/claude-opus-4-6) | Anthropic | **59.3** | 57.8 | 63.0 | $5 | $25 | | 25 | [DeepSeek V4 Pro](https://noometry.com/models/deepseek-v4-pro) (open weights) | DeepSeek | **59.0** | 56.5 | 64.8 | $0.66 | $1.98 | Sponsored placements are available on pages like this one. [Advertise on Noometry](https://noometry.com/advertise) Top two head to head: [GPT-6 Astra vs GPT-6.1 Sol](https://noometry.com/compare/gpt-6-1-sol-vs-gpt-6-astra) ## Frequently asked questions ### What is the best ai model for hard reasoning? For reasoning, GPT-6 Astra scores highest on the current data (87.6), weighting reasoning 70%, math 30%. The cheapest model in the top 10 is GPT-6.1 Sol at $2 / $10 per million tokens. ### How is this shortlist built? Novel problem solving is measured best by abstraction puzzles and math. Each model's category scores are blended with those weights; only ranked models with results in every needed category are listed. ## Other jobs - [Best AI model for coding](https://noometry.com/best/for/coding) - [Best AI model for debugging](https://noometry.com/best/for/debugging) - [Best AI model for building websites](https://noometry.com/best/for/frontend) - [Best local LLM for coding](https://noometry.com/best/for/local-code) - [Best AI model for agents](https://noometry.com/best/for/agents) - [Best AI model for RAG](https://noometry.com/best/for/rag) - [Best AI model for research](https://noometry.com/best/for/research) - [Best AI model for studying](https://noometry.com/best/for/studying) - [Best AI model for math](https://noometry.com/best/for/math) - [Best AI model for data analysis](https://noometry.com/best/for/data-analysis) - [Best AI model for Excel](https://noometry.com/best/for/spreadsheets) - [Best AI model for data extraction](https://noometry.com/best/for/extraction) - [Best AI model for writing](https://noometry.com/best/for/writing) - [Best AI model for emails](https://noometry.com/best/for/emails) - [Best AI model for copywriting](https://noometry.com/best/for/marketing) - [Best AI model for creative writing](https://noometry.com/best/for/fiction) - [Best AI model for translation](https://noometry.com/best/for/translation) - [Best AI model for language learning](https://noometry.com/best/for/language-learning) - [Best AI model for customer service](https://noometry.com/best/for/support) - [Best AI model for presentations](https://noometry.com/best/for/presentations) - [Best AI model for PDFs](https://noometry.com/best/for/documents) - [Best AI assistant for everyday use](https://noometry.com/best/for/general) ### Cite this page Noometry. (2026). Best AI model for hard reasoning. Retrieved October 10, 2026, from https://noometry.com/best/for/reasoning Quote Noometry with a link back to this page. It is also available in [Markdown](https://noometry.com/md/best/for/reasoning.md).