Find your next model.
Published prices, capabilities and lifecycle notices. One clear comparison.
Flagship starting set · Latest eligible release from each reviewed family · Checked twice daily for new visits. Edit any column to make it yours.
A few things to know
How is the starting comparison chosen?
The starting set uses the newest published release within reviewed OpenAI Astra, Anthropic Opus and Google Pro/Argon flagship families, with one direct-provider model per family. Preview, specialist, future-dated and deprecated models are excluded, and published text prices are required. New family names require review. This is a starting set, not a benchmark ranking or a claim of equal performance. Your shared selections stay unchanged.
What can I compare?
Compare up to four exact provider model IDs: published input, output and cached text-token rates, context windows, maximum output, tool calling, structured output, reasoning and input modalities.
Where does the comparison data come from?
Pricing and capability metadata come from Models.dev and are refreshed on the first request after 00:00 and 12:00 UTC per server process. Lifecycle notices and replacement recommendations come from ModelClock’s published official-provider evidence. Each column links to its sources and shows when metadata was retrieved.
Is this my actual API bill?
No. The monthly estimate uses the text-token volumes and cached-input share you enter. It is not connected to your billing account. Special pricing tiers, long-context surcharges, cache writes, images, audio, tools and tax are excluded. Include billed reasoning tokens in output. Missing required prices remain unknown.
Does a recommended replacement work without code changes?
Not necessarily. Recommendations reproduce the provider’s published notice for that hosting route. Metadata does not guarantee availability, output quality or API compatibility. Test your workload before switching; ModelClock does not run paid benchmarks or rank one model as universally best.