OctoRouterModels
Learn how Octo Router uses local embeddings to route requests based on intent.
The section discusses the models section of the configuration field
Defaults
A default model & maxTokens can be provided for each of the provider
models:
defaults:
openai:
model: "openai/gpt-4o-mini"
maxTokens: 4096
anthropic:
model: "anthropic/claude-haiku-3"
maxTokens: 4096
gemini:
model: "gemini/gemini-2.5-flash-lite"
maxTokens: 4096The defaults must be specified using the same name with it's respective provider.
Catalog
OctoRouter has a built in catalog of models, but you may want to add new models or override existing ones.
models:
catalog:
- id: "openai/gpt-4o"
provider: "openai"
name: "GPT-4o (Special Rate)"
inputCost: 2.00
outputCost: 8.00
contextWindow: 128000
tier: "premium"
capabilities:
- "coding"
- "reasoning"
- "creative"inputCost and outputCost are priced per 1,000,000 tokens