Model Catalog

Read from your connected providers, not typed in. Everything a provider publishes is synced; the one field none of them publish is price.

Add models
Choose a provider and pick from what it actually offers — name, model id and context window come straight from the provider, so there is nothing to type and nothing to mistype.
Pick a provider above to see everything it offers.
Drift from your providers
Check what each provider publishes against what the catalog says. Nothing is written until you choose it.
Models with access
10 we currently have API access to
ModelProviderBest forContextCost (in/out per M)AccessStatusRouted
Claude Fable 5AnthropicHighest-capability tier -- justify with an eval before using over Opus1,000,000$10 / $50
Have access
EvaluatingNot routed
Claude Haiku 4.5AnthropicClassification, extraction, routing, short summaries200,000$1 / $5
Have access
EvaluatingIn use
Claude Opus 4.8AnthropicComplex reasoning and long-horizon agentic coding where frontier quality matters more than cost1,000,000$5 / $25
Have access
EvaluatingNot routed
Claude Opus 5AnthropicComplex agentic coding and enterprise-grade work1,000,000$5 / $25
Have access
EvaluatingNot routed
Claude Sonnet 5AnthropicFast, general-purpose applications (promo pricing through Aug 2026)1,000,000$2 / $10
Have access
EvaluatingIn use
Gemini 3.5 Flash-LiteGoogleClassification, extraction, tagging, short summaries1,000,000$0.3 / $2.5
Have access
EvaluatingNot routed
Gemini 3.6 FlashGoogleBalanced general-purpose Flash tier1,000,000$1.5 / $7.5
Have access
EvaluatingNot routed
GPT-5.6 LunaOpenAIClassification, extraction, routing1,000,000$1 / $6
Have access
EvaluatingNot routed
GPT-5.6 SolOpenAIFrontier coding, reasoning, and agentic work1,000,000$5 / $30
Have access
EvaluatingNot routed
GPT-5.6 TerraOpenAIProduction-default general-purpose tasks1,000,000$2.5 / $15
Have access
EvaluatingNot routed
Reference library
1 tracked for awareness. Flip Access on and Atlas makes one tiny real call to check — it only turns on if the provider actually answers.
ModelProviderBest forContextCost (in/out per M)AccessStatusRouted
Gemini 3.1 ProGoogleHard reasoning and agentic coding1,000,000$2 / $12
Reference only
EvaluatingNot routed