Routing

Choose the model a request uses by default, the order of fallbacks, and how the gateway reacts to upstream failure.

Default model

Used whenever a request does not pin a model of its own.

Provider: Anthropic
Model ID: claude-sonnet-5
Context: 1M

Routing strategy

The first healthy model in the chain serves the request.

Fallback chain

Tried in order when the default model fails, times out or is rate limited.

3 configured

1GLM 5.3 Flashglm-5.3-flash
2DeepSeek V4.1 Flashdeepseek-v4.1-flash
3GLM 5.2glm-5.2

Add to the chain

Behaviour

Applied to every request that goes through the chain.

Retry failed requestsRetry once before moving down the fallback chain.Skip unavailable modelsBypass models reporting degraded upstream health.Prefer lowest latencyFavour the fastest healthy model in the chain.Prefer lowest costFavour the cheapest healthy model in the chain.

Request path

How a request moves through the gateway with this configuration.

1Claude Sonnet 5claude-sonnet-5Default
2GLM 5.3 Flashglm-5.3-flashFallback
3DeepSeek V4.1 Flashdeepseek-v4.1-flashFallback
4GLM 5.2glm-5.2Fallback

Provider health

Live upstream signal used by automatic routing.

Claude Sonnet 5claude-sonnet-5
1,285 msHealthy
GLM 5.3 Flashglm-5.3-flash
1,284 msHealthy
GLM 5.2glm-5.2
1,326 msHealthy
DeepSeek V4 Prodeepseek-v4-pro
1,182 msDegraded
DeepSeek V4.1 Flashdeepseek-v4.1-flash
597 msHealthy
GPT 5.6 Lunagpt-5.6-luna
1,291 msHealthy