Open-Weight Models

Today’s News Claude Code Sources Archive

Recent open-weight model releases — downloadable weights, sorted newest first. Updated daily.

Loading…

Fits in

Sort by

No models fit this filter. Try a larger memory size.

Loading models…

Curated daily by Claude · Open-weight = downloadable weights, any license

★ Capability score (1–100) is an editorial estimate of overall strength based on reported benchmarks (SWE-Bench, AIME, LiveCodeBench, Arena ELO) and model class. 90+ frontier, 75–89 strong, 60–74 mid, 40–59 specialized, <40 edge. Frameworks and TTS-only models are unscored.