Best LLM for every task

47 task rankings, each computed from public benchmarks and practitioner reports and updated as the evidence changes. Sub-tasks sit under the task they narrow. See the methodology for how scores are computed.

Best LLM for Coding

Current #1: Claude Fable 5.2

Best LLM for AI Agents

Current #1: Claude Fable 5.2

Best LLM for Writing

Current #1: Gemini 3.8 Flash

Best LLM for RAG & Long Context

Current #1: Gemini 4 Carbon

Best LLM for Chat & Roleplay

Current #1: Gemini 3.8 Flash

Best LLM for Translation

Current #1: Gemini 3.8 Flash

Best LLM for Summarization

Current #1: Gemini 4 Carbon

Best LLM for Structured Data Extraction

Current #1: DeepSeek V4.1 Flash

Best LLM for Math & Reasoning

Current #1: Claude Opus 5.5

Best Cheap LLM for High Volume

Current #1: GLM 5.2

Best LLM for Vision & Documents

Current #1: Gemini 3.8 Flash

Best Decision Model for Agent Routing and Scoring

Current #1: H2O-Lightning-4B v1.1 (H2O.ai, Qwen3.5-4B fine-tune, stock vLLM + open shim)