Transparent model data instead of marketing claims
Model cards, a quant and metadata registry with SHA hashes and file manifests, and reproducible benchmark runs. Choose models for your tasks based on data, not marketing.
A catalog and benchmarks with verifiable artifacts
Description, provenance, licenses and key parameters for every model
Versions, SHA hashes and file manifests for exact artifact matching
Complete file listings for each model repository to verify integrity
Reproducible test runs with pinned configurations
A single catalog for comparing models before adopting them in a product
Benchmark results are published together with the conditions they were measured under
Search the public catalog and model cards
SHA hashes, file manifests and quant metadata
Benchmark runs with open configurations
A well-founded model choice for your task and hardware
Tell us about your task — let's discuss the AIRuns catalog and benchmarks