autoCapability-matched selection: a required-tier floor keeps only capable models, then a shrinkage-blended learned reward picks the cheapest of them. General workloads.
Model catalog
40 curated models. 9 providers.
One base_url.
Hand-curated catalog from Anthropic, OpenAI, Google, x-AI, DeepSeek,
Perplexity, Moonshot, Z.ai, and Groq — led by the Claude 5 family
(Opus 5, Sonnet 5, Fable 5). OpenAI- and Anthropic-compatible endpoints.
Set model="auto" and capability-matched
selection picks the best one per request.
Catalog
9 providers, ranked on merit.
Per-model quality, latency, and success variance are tracked live from real traffic. The leaderboard is public to every caller via GET /v1/intelligence/rankings — and GET /v1/catalog/runnable shows only what will actually route right now.
Anthropic
Claude 5 · frontierclaude-opus-5 · claude-sonnet-5 · claude-fable-5 · claude-opus-4-7 · claude-haiku-4-5
OpenAI
GPT-5 · reasoninggpt-5.6-sol · gpt-5.5 · gpt-4.1 · gpt-4o · o3 · o4-mini
gemini-3.5-flash · gemini-3.1-flash-lite · gemini-2.5-pro · gemini-2.5-flash
x-AI
Grok 4.xgrok-4.6 · grok-4.5 · grok-4.3 · grok-4.20
DeepSeek
V4deepseek-v4-pro · deepseek-v4-flash
Perplexity
search-augmentedsonar-pro · sonar-reasoning-pro · sonar-deep-research · sonar
Moonshot
Kimikimi-k3 · kimi-k2.6 · kimi-k2.5
Z.ai
GLMglm-5.2 · glm-5
Groq
fast inferenceopenai/gpt-oss-120b
Embeddings
OpenAItext-embedding-3-small · text-embedding-3-large
Auto-routing
Set model to a strategy, not a name.
auto:fastLowest p50 latency model. Real-time UX, streaming interfaces.
auto:floorCheapest model above the quality threshold. Bulk processing, classification, triage.
auto:bestHighest quality regardless of cost. Critical reasoning, code review, legal.
Prometheus stress-test
Simple tasks routed to DeepSeek at $0.000005/req. Complex reasoning routed to Claude Opus at $0.000156/req. A 30×+ cost spread, chosen automatically — per request, from the learned cost/quality frontier, not hard-coded rules.