Docs

OpenDoor lists open-weight models first: Qwen, DeepSeek, Llama, Mistral, Gemma from Hugging Face and Ollama, plus optional closed providers.

Model Catalog Overview

OpenDoor lists open-weight models first — Qwen, DeepSeek, Llama, Mistral, Gemma — from Hugging Face and Ollama, plus optional closed providers.

You do not wait for Azure AI Foundry to publish a model. Run:

bash
bun run ingest:open-models

New HF repos for tracked orgs (Qwen, deepseek-ai, meta-llama, mistralai, moonshotai, …) appear as warming the same day. Models we can already serve (vendor API or known Ollama tag) are live or dedicated.

Statuses

Status	Meaning
Live	Call `POST /v1/chat/completions` now
Warming	In catalog; serving path coming shortly
Needs GPU (dedicated)	Deploy from Dashboard → Deployments (this Mac or GCP)
On request	Closed marketplace models only

Browse Models

bash
curl http://localhost:3001/v1/models \
  -H "Authorization: Bearer YOUR_API_KEY"

In the dashboard: Models — filter Open weight, Live, or New this week. Prices are shown as USD per 1M tokens.

Pricing posture

Open-weight models use a low markup (about 5–15%). Closed Western models use a higher markup (~20%). We do not surcharge Chinese open models.

Next Steps

View Live Models
Available on Request (closed models)

Premium rentals

Live models