Why this matters
A source-backed model catalog matters because the AI release landscape is noisy. Labs announce models with overlapping names, varying access terms, and inconsistent capability claims. By pinning each record to Epoch AI's CC-BY catalog and recording the extraction path, Model Gauntlet gives readers a verifiable starting point for comparing releases without trusting a single press release.
For operators choosing a model, the catalog answers foundational questions: when was it released, who made it, how can it be accessed, and are the weights available? Those facts are prerequisites for any benchmark or price comparison.
What entered the model guide
The reviewed set covers 22 tracked labs. Each record preserves its model name, lab, release date, domains, tasks, access classification, open-weight status, primary source links, and extraction path.
The catalog gives readers a source-backed way to browse releases and operating characteristics without turning catalog facts into a performance ranking.
What the catalog does not prove
Inclusion is not an endorsement. A catalog record does not establish benchmark leadership, adoption, price, latency, reliability, or overall superiority.
Model Gauntlet keeps Epoch catalog facts separate from evaluation results. A model only receives a current LMArena value when the published provider-frontier snapshot contains an exact matched record.
What we know
The catalog was retrieved from Epoch AI on July 21, 2026 and is licensed CC-BY.
Records include primary source URLs so readers can inspect the original row.
Each model is matched to a tracked lab and normalized domains such as language, vision, audio, and multimodal.
What we don't know
Pricing, context-window length, and enterprise terms are not in the Epoch source, so they are not included in these records.
Benchmark leadership and adoption are tracked separately through LMArena and other approved sources.
Model lifecycle changes after release, such as deprecation or version updates, require a separate provider-catalog observation.
Direct answers
Frequently asked questions
What is the Model Gauntlet model guide?
It is a browsable catalog of source-backed AI model records. Each entry preserves release date, lab, domains, tasks, access type, and open-weight status from Epoch AI's Data on AI Models catalog.
How many models are in the guide?
The reviewed set currently contains 456 model records across 22 labs.
Does inclusion mean a model is recommended?
No. Inclusion only means the model is documented in the reviewed source. It does not imply benchmark leadership, price competitiveness, or suitability for any specific use case.
Where do the model records come from?
Records are derived from Epoch AI's Data on AI Models catalog, retrieved July 21, 2026 under a CC-BY license.