Useful slices of the catalog
All reportsLLM release reports
A small set of data-backed reports built from the release catalog. These pages are meant to answer real recurring questions, not multiply near-duplicate search pages.
Report
Release radar
A concise view of the freshest releases, updates, retirements, and source links.
Report
Frontier model leaderboard
Frontier-class models ranked by their strongest tracked benchmark, with context, price, and access.
Report
LLM API pricing
Models with published API pricing, sorted cheapest-first by input and output cost per million tokens.
Report
Model access matrix
API, open-weight, preview, and retired models grouped by how teams can actually use them.
Report
Open coding models
Open-weight coding models compared by release date, context, size, and available SWE-bench claims.
Report
Local LLM shortlist
Downloadable models with compact total or active parameter counts and rough local memory estimates.
Report
Release calendar
A month-by-month release history grouped by lab and event type.
Evergreen guides
Reference pages for reading the catalog
Guide
What is an open-weight model?
How LLM Releases distinguishes proprietary, open-weight, and open-source model releases.
Guide
How to size local LLM hardware
A practical explanation of parameters, quantization, context windows, and rough memory estimates.
Guide
How to read model release timelines
Use lab, family, and changelog pages to follow model lineages without losing source provenance.
Guide
LLM benchmarks explained
What MMLU, GPQA, SWE-bench, AIME, LMArena and other evals actually measure — and why scores need a source.
Guide
What is a reasoning model?
How reasoning or “thinking” models differ from standard chat models, and when the extra latency is worth it.