{"name":"ossmodeldb","description":"Exact memory and modeled speed for locally-run open-weight models. Weights are summed from published file bytes; KV cache is computed per layer.","license":"https://creativecommons.org/licenses/by/4.0/","attribution":"ossmodeldb.org — CC BY 4.0","counts":{"models":"2382","artifacts":"67146","accelerators":"239"},"endpoints":{"GET /api/v1/models":"list base models; ?modality=&limit=","GET /api/v1/models/{slug}":"one model with architecture and every shipped quantization","GET /api/v1/fit":"fit a model to hardware; ?model=&hw=&ctx=&kv=","GET /api/gguf":"parse any Hub GGUF header live; ?url="},"notes":["Byte counts are exact and summed across shards, never params x bits-per-weight.","KV figures assume default configuration; disabling sliding-window attention changes them.","Speed is modeled, not measured. Every speed carries an error band."]}