Skip to main content

Models

Use documented model IDs in requests. Aliases are accepted for common model names, but exact IDs are best for automation and reproducible routing.

Claude, Opus 5, and Fable 5

Claude models use Anthropic-compatible routes. Opus 5 follows the current Opus routing policy, while Fable 5 uses its Anthropic-compatible ID and OpenAI-backed route.

ModelFamilyEndpoints
anthropic/claude-opus-5Opus 5/v1/messages, /v1/chat/completions
anthropic/claude-opus-4.8Opus 4.8/v1/messages, /v1/chat/completions
anthropic/claude-opus-4.7Opus 4.7/v1/messages, /v1/chat/completions
anthropic/claude-sonnet-5Sonnet 5/v1/messages, /v1/chat/completions
anthropic/claude-haiku-4.5Haiku 4.5/v1/messages, /v1/chat/completions
anthropic/fable-5Fable 5/v1/messages, /v1/chat/completions, /v1/responses

GPT

GPT models use OpenAI provider routes and are available through OpenAI-compatible endpoints.

ModelFamilyRouting
openai/gpt-5.6-solGPT-5.6 SolUpstream gpt-5.6-sol
openai/gpt-5.6-terraGPT-5.6 TerraUpstream gpt-5.6-terra
openai/gpt-5.6-lunaGPT-5.6 LunaUpstream gpt-5.6-luna
openai/gpt-5.5GPT-5.5Upstream gpt-5.5
openai/gpt-5.4GPT-5.4Upstream gpt-5.4
openai/gpt-5.4-miniGPT-5.4 MiniUpstream gpt-5.4-mini

Open model groups

Open model-group IDs route through the configured external/open provider paths for the matching route.

ModelFamily
z-ai/glm-5.3GLM 5.3
z-ai/glm-5.2GLM 5.2
z-ai/glm-5.1GLM 5.1
z-ai/glm-5-turboGLM 5 Turbo
minimax/minimax-m3MiniMax M3
deepseek/deepseek-v4-proDeepSeek V4 Pro
deepseek/deepseek-v4-flashDeepSeek V4 Flash

Endpoint compatibility

EndpointSupported families
/v1/messagesClaude, Opus 5, Fable 5, open model groups
/v1/chat/completionsClaude, Opus 5, Fable 5, GPT, open model groups
/v1/responsesFable 5, GPT, open model groups
/v1/modelsAll documented models available to the authenticated key

Send the first request.

Create a key, choose a compatible endpoint, and use the SDK you already have.