Skip to content
All systems operationalStatus

Models & protocols

Model IDs, tiers, and which endpoint serves which family.

Model IDs

Send one of these IDs as model. IDs are case-sensitive. You can also fetch the live list from GET /v1/models, or browse rates on the Models page.

Model IDFamilyTierContext
claude-fable-5-1ClaudeFrontier262K
claude-fable-5ClaudeFrontier262K
claude-mythos-5ClaudeFrontier262K
claude-opus-5.5ClaudeFrontier262K
claude-opus-5ClaudeFrontier262K
claude-opus-4.8ClaudeFrontier262K
claude-opus-4.7ClaudeFrontier262K
claude-opus-4.6ClaudeFrontier262K
claude-opus-4.5ClaudeFrontier262K
claude-opus-4.1ClaudeFrontier262K
claude-opus-4ClaudeFrontier262K
claude-sonnet-5ClaudeBalanced131K
claude-sonnet-4.6ClaudeBalanced131K
claude-sonnet-4.5ClaudeBalanced131K
claude-sonnet-4ClaudeBalanced131K
claude-haiku-4.5ClaudeSpeed131K
claude-haiku-3.5ClaudeSpeed131K
gpt-astra-6ChatGPTFrontier262K
gpt-5ChatGPTFrontier262K
gpt-5-miniChatGPTBalanced131K
gpt-4.1ChatGPTBalanced131K
gpt-4.1-miniChatGPTSpeed131K
gpt-4oChatGPTBalanced131K
gpt-4o-miniChatGPTSpeed131K

Protocol support

Claude models are native to the Anthropic Messages API; GPT models are native to the OpenAI Responses API. Chat Completions works with every model, which is what most third-party tools use.

EndpointClaude modelsGPT models
POST /v1/messagesYesNo
POST /v1/responsesNoYes
POST /v1/chat/completionsYesYes
GET /v1/modelsYesYes

Choosing a tier

Every model in a tier bills at the same rate, so choose by capability:

  • Frontier

    The most capable models, for deep reasoning and long-horizon agent work.

    e.g. claude-fable-5-1, claude-mythos-5, claude-opus-5.5, gpt-astra-6, gpt-5

  • Balanced

    Strong everyday models that trade a little depth for speed and cost.

    e.g. claude-sonnet-5, gpt-5-mini

  • Speed

    Fast, lightweight models for high-volume and latency-sensitive work.

    e.g. claude-haiku-4.5

A common pattern: plan and review with a Frontier model, do the bulk of the work with a Balanced one, and fan out sub-agents or classification on a Speed model. Claude and ChatGPT models can be mixed freely.

Simple prompts may be served by a lighter model of the same family. See How routing works.