The best model for each job
The leaderboard, cut by what you are actually trying to do. Every ranking is derived from the same sourced data and refreshes when Wren updates the table. Updated Oct 10, 2026.
- Best AI model for coding#1 Claude Opus 5.5 · 3 models
Which model to put behind a coding assistant or agentic coding workflow, ranked by independent evals, price and what engineers report in production.
- Best AI model for agents#1 Claude Opus 5.5 · 4 models
Models that hold up over long tool-using runs: context, reliability, price per million tokens and what teams running agents in production say.
- Best AI model for customer support and chat#1 Gemini 4 Argon · 6 models
Models for customer-facing assistants: chat quality on Arena, latency, price at volume and the vendors enterprises are actually deploying.
- Best AI model for extraction and classification#1 Claude Haiku 5.5 · 3 models
Small, fast models for structured extraction, classification and routing tiers, with the price cliffs that matter at scale.
- Best AI model for hard reasoning#1 GPT-6 Astra · 3 models
The frontier models for research-grade analysis and problems the workhorse tier fails on, with the independent exam scores and what they cost.
- Cheapest AI models that are still good#1 Claude Haiku 5.5 · 12 models
Every model on the leaderboard ordered by blended price per million tokens, with its independent intelligence score, so you can see what each dollar buys.