Every model. One key. One bill.
89 models across every major lab — Claude, GPT-5, Gemini 3, DeepSeek, Groq, Qwen, Kimi — behind a single OpenAI-compatible endpoint. Route each turn to the model that fits, and skip the flat per-seat subscription entirely.
# one endpoint, one key from `th api login`
base_url = "https://llm.smoo.ai/v1"
model = "groq-gpt-oss-120b" # cheap default
model = "claude-opus-4-8" # frontier on demandPay for the tokens you use — not a premium seat.
Frontier coding assistants bill a flat monthly seat or lock you to one lab's per-token rate. The gateway does neither: it's metered per org across every model, so the economical move is the default one.
Cheap by default
Send the heavy lifting to fast open models — groq-gpt-oss-120b, deepseek-v4-flash, qwen3-coder-flash — at a fraction of frontier cost. Reach for claude-opus-4-8 or gpt-5.5-pro only when the task earns it.
No per-seat lock-in
One org key, billed by token. No monthly minimum per developer, no paying for a seat that sat idle over the weekend — the opposite of a flat coding subscription.
Switch models, not code
It speaks the OpenAI API. Change one string to move from a frontier model to an open one — same SDK, same request shape, no re-integration.
Every major lab, one roster.
Query it live any time with th cast models — the roster moves as labs ship.
Anthropic
claude-opus-4-8 · sonnet-4-6 · haiku-4-5
OpenAI
gpt-5.5-pro · gpt-5.2-codex · o3 · gpt-4.1
gemini-3.1-pro-preview · 3.5-flash · flash-lite
DeepSeek
deepseek-v4-pro · v4-flash · r1 · reasoner
Groq
groq-gpt-oss-120b · oss-20b · qwen3.6-27b
Qwen
qwen-3.7-max · qwen3-coder-flash · qwen-plus
Moonshot
kimi-k2.6 · k2.5 · k2-thinking
MiniMax
minimax-m2.7 · m2.5-highspeed · m2.1
Zhipu & more
glm-5.1 · glm-5-turbo · mimo-v2.5-pro
The engine under everything Smooth.
Big Smooth & the operatives
Every agent and the orchestrator resolve their model slots through the gateway — a fast mode drops the fleet to a cheap open model for quick turns.
Your own apps, same key
Org-scoped by JWT from th api login. Point any OpenAI SDK at llm.smoo.ai/v1and you're on the same metered pool as the platform.