Best LLMs, by the number that matters to you
Top three: overall · read October 20
- Cheapest LLM per solved coding task
- Best open-weight LLM for coding
- Best LLM for web development
- Cheapest good LLM for coding
- LLMs with the longest context window
- Cheapest LLM API
88 lists, each one a filter and an order over the same dataset, refreshed with it every day. 75 have five or more models and a number that separates them; the rest are kept for reference.
- Best LLM under $1 per million tokens Xiaomi MiMo-V2.6-Pro 46.3
- Best LLM under $5 per million tokens Meta Muse Spark 1.3 48.1
- Best LLM under $15 per million tokens Anthropic Claude Sonnet 5.5 56.0
- Best open-weight LLM overall
- 1 Xiaomi MiMo-V2.6-Pro 46.3
- 2 Z.ai GLM 5.3 44.8
- 3 StepFun Step 5 Preview 43.7
- Best LLM for images (vision) Anthropic Claude Opus 5.5 57.6
- Best LLM for PDFs and files Anthropic Claude Opus 5.5 57.6
- Best LLM with audio input Google Gemini 4 Argon 52.6
- Best LLM for tool calling Anthropic Claude Opus 5.5 57.6
- Best LLM for structured outputs (JSON) Anthropic Claude Opus 5.5 57.6
- Best reasoning model Anthropic Claude Opus 5.5 57.6
- Best LLM with a 1M token context window
- 1 Anthropic Claude Opus 5.5 57.6
- 2 Anthropic Claude Sonnet 5.5 56.0
- 3 Anthropic Claude Fable 5.1 53.4
- Best LLM for terminal and DevOps tasks Anthropic Claude Opus 5.5 65.2%
- Best LLM for competitive programming Anthropic Claude Fable 5.1 90.5%
- Best LLM for olympiad programming (IOI) Google Gemini 4 Argon 100.0%
- Best LLM for fixing GitHub issues
- 1 Anthropic Claude Opus 5 97.0%
- 2 DeepSeek V4 Pro 0813 96.4%
- 3 OpenAI GPT-5.6 Sol 96.2%
- Best LLM for code migration Anthropic Claude Sonnet 5.5 69.8%
- Best LLM on Vibe Code Bench Anthropic Claude Sonnet 5.5 92.4%
- Best LLM for computer use Anthropic Claude Opus 5 31.4%
- Best LLM for long autonomous tasks Anthropic Claude Opus 4.6 78.9%
- Best LLM for research tasks Anthropic Claude Opus 5 88.9%
- Best LLM for web browsing agents Anthropic Claude Fable 5.1 100.0%
- Best LLM for data analysis
- 1 Google Gemini 4 Argon 85.2%
- 2 OpenAI GPT-6 Astra 83.0%
- 3 OpenAI GPT-6.1 Sol 82.7%
- Best LLM for research-level math OpenAI GPT-6 Astra 93.7%
- Best LLM for math proofs Anthropic Claude Fable 5.1 100.0%
- Best LLM on Humanity's Last Exam OpenAI GPT-6 Astra 54.8%
- Best LLM for abstract reasoning (ARC-AGI-2) OpenAI GPT-6 Astra 95.0%
- Most factually accurate LLM Google Gemini 4 Argon 76.0%
- Best LLM for creative writing Anthropic Claude Opus 5.5 1534
- Best LLM for chat
- 1 Google Gemini 4 Argon 1532
- 2 Anthropic Claude Opus 5.5 1516
- 3 Anthropic Claude Fable 5.1 1512
- Best LLM for following instructions Anthropic Claude Opus 5.5 1536
- Best LLM for understanding images Anthropic Claude Fable 5 1325
- Best LLM for UI design OpenAI GPT-6 Astra 1400
- Best LLM for game development OpenAI GPT-6 Astra 1474
- Best LLM for data visualization Anthropic Claude Opus 5.5 1397
- Best LLM for 3D scenes OpenAI GPT-6 Astra 1505
- Best LLM for professional work (GDPval)
- 1 OpenAI GPT-5.2 49.7%
- 2 Anthropic Claude Opus 4.5 45.5%
- 3 Anthropic Claude Opus 4.1 43.6%
- Best LLM for freelance projects Anthropic Claude Opus 5.5 21.3%
- AA Intelligence Index Intelligence Claude Opus 5.5 57.6
- AA Agentic Index Agentic work Claude Fable 5.1 57.9
- GPQA Diamond Reasoning Gemini 3.1 Pro Preview 95.5%
- AIME Math Gemini 3.1 Pro Preview 98.1%
- MMLU Pro Knowledge Claude Fable 5.1 92.4%
- Design Arena: websites Web design Muse Wonder 1360
- Vals Index General Claude Sonnet 5.5 70.8%
- Cheapest good LLM for overall intelligence DeepSeek V4.1 Flash $0.0112
- Best open-weight LLM for overall intelligence Xiaomi MiMo-V2.6-Pro 46.3
- Cheapest good LLM for agentic work inclusionAI Ling 3.0 Flash $0.063
- Best open-weight LLM for agentic work Z.ai GLM 5.3 53.1
- Cheapest good LLM for reasoning inclusionAI Ling 3.0 Flash $0.063
- Best open-weight LLM for reasoning MoonshotAI Kimi K3 92.9%
- Cheapest good LLM for math Qwen3.5-Flash $0.26
- Best open-weight LLM for math MoonshotAI Kimi K2.5 95.6%
- Cheapest good LLM for knowledge Z.ai GLM 5.3 Flash $0.16
- Best open-weight LLM for knowledge MoonshotAI Kimi K3 88.0%
- Cheapest good LLM for web design DeepSeek V4.1 Flash $0.0112
- Best open-weight LLM for web design MoonshotAI Kimi K3 1335
- Cheapest good LLM for general-purpose work Anthropic Claude Haiku 5.5 $0.50
- Best open-weight LLM for general-purpose work DeepSeek V4.1 Flash 1017 61.3%
- Best OpenAI model overall OpenAI GPT-6 Astra 52.7
- Best Qwen model overall Qwen3.8 Max (0902) 45.4
- Best Google model overall Google Gemini 4 Argon 52.6
- Best Anthropic model overall
- 1 Anthropic Claude Opus 5.5 57.6
- 2 Anthropic Claude Sonnet 5.5 56.0
- 3 Anthropic Claude Fable 5.1 53.4
- Best Mistral AI model overall Mistral Large 4 38.4
- Best Z.ai model overall Z.ai GLM 5.3 44.8
- Best DeepSeek model overall DeepSeek V4.1 Flash 1017 39.5
- Best Meta Llama model overall Meta Llama 4 Maverick 9.3
- Best Moonshot AI model overall MoonshotAI Kimi K3 43.6
- Best MiniMax model overall MiniMax M3 29.2
Best model from each maker
Every maker with at least one scored model. Lists under five models stay out of search results.
- Best OpenAI model OpenAI GPT-6 Astra, index 88.7 33
- Best Anthropic Claude model Anthropic Claude Opus 5.5, index 92.2 15
- Best Google Gemini model Google Gemini 2.5 Pro Preview 06-05, index 83.1 15
- Best Qwen model Qwen3.8 Flash, index 87.4 14
- Best DeepSeek model DeepSeek V4.1 Flash 1017, index 83.3 11
- Best Z.ai model Z.ai GLM 5.3, index 73.1 9
- Best Moonshot AI model MoonshotAI Kimi K2 0711, index 64.7 6
- Best Meta model Meta Muse Spark 1.3, index 79.7 5
- Best MiniMax model MiniMax M2.5, index 59.4 4
- Best xAI model SpaceXAI Grok 4.6, index 76.4 4
- Best Xiaomi model Xiaomi MiMo-V2.6-Pro, index 72.6 4
- Best Mistral AI model Mistral Large 4, index 55.3 3
- Best Tencent model Tencent Hy4 preview, index 61.6 3
- Best Meta Llama model Meta Llama 4 Maverick, index 18.1 2
- T Best Thinking Machines model Thinking Machines Inkling, index 49.7 2
- Best Arcee AI model Arcee AI Trinity Large Thinking, index 5.4 1
- Best Inception model Inception Mercury 2, index 0.0 1
- Best NVIDIA model NVIDIA Nemotron 3.5 Super VL, index 65.5 1
- R Best Reflection model Reflection Beam 1.0 501B A23B, index 33.8 1
- Best StepFun model StepFun Step 5 Preview, index 71.4 1
- Best Upstage model Upstage Solar Pro 4, index 24.4 1