LLM directory

Thousands of language models exist if you count every fine-tune on Hugging Face. This directory tracks the models most teams actually evaluate: current and recent releases from the major API providers and the leading open-weight labs, with prices and specs checked against each provider's documentation.

Blended price combines input and output rates into one number using a three-to-one ratio of input to output tokens, which is typical of chat and retrieval apps. It's a quick way to compare models; use the cost calculator for your real token mix.

Tiers group models by how their providers position them and how they place on public benchmarks: frontier for the most capable models, advanced for strong mid-priced models, efficient for fast, low-cost models and compact for small models that run on modest hardware. Tiers are our editorial grouping, not our own test results.

Some open-weight models aren't sold per token by a major host, and some new or preview models don't have a confirmed list price yet. We only show prices we could confirm on an official pricing page.