Best AI for Coding
Ranked by SWE-bench Verified — real GitHub issues resolved end-to-end — with League Score as the tiebreak.
See the ranking →Task rankings
Best AI for the job.
Task-first rankings generated from the League database — each list states the metric it is ranked by.
Best AI for Reasoning
Ranked by GPQA Diamond, the graduate-level science benchmark built to resist lookup.
See the ranking →Best AI for Research
Research work rewards hard reasoning and long context: ranked by GPQA Diamond, then League Score.
See the ranking →Best AI for Writing
Ranked by crowd preference (LMArena Elo) where available — writers are the toughest crowd — then League Score.
See the ranking →Best AI for Image Understanding
Only models that accept image input qualify; ranked by League Score.
See the ranking →Best AI for Budget Chat
Chat volume is a cost game: ranked by League Score per dollar (Value Division).
See the ranking →Best AI for Long Documents
Ranked by context window first — the document has to fit before quality matters — then League Score.
See the ranking →Best AI for Math
Ranked by MMLU-Pro — the only benchmark we track with dedicated mathematics subject categories — across every model it has measured.
See the ranking →Best AI Chatbot
Ranked by LMArena Elo: blind head-to-head human votes at scale, the best public proxy for how a chatbot actually feels to talk to.
See the ranking →Best Open-Source AI Models
Every open-weights model in the database, ranked by League Score — models you can download, self-host and fine-tune.
See the ranking →Fastest AI Models
Ranked by measured output speed (tokens per second). Only models with a sourced speed measurement appear — coverage is stated, never padded.
See the ranking →Best AI for Data Analysis
Ranked by combined coding and reasoning League positions: data work is code plus inference, so both rankings count.
See the ranking →Best AI for Summarization
Models with at least a 500K-token context window, ranked by blended price — summarization is long documents at volume, so the window qualifies them and cost orders them.
See the ranking →Best AI for Customer Support
Models at or above a League Score of 75, ranked by blended price — a support bot answers all day, so the quality floor comes first and cost per answer orders the list.
See the ranking →