Your models, in your editor
One key works everywhere: Claude Code, Codex, and anything that speaks the OpenAI or Anthropic API. Open weights, per-token pricing, and nothing kept after the answer.
- Create a key at turboprivate.ai → Settings → Connect your IDE. It is shown once.
- Run one line.
npx turboprivate@latest login npx turboprivate@latest claude # or: codex - Work. The CLI ends with a real request, so you know it is connected — not a guess.
What we serve right now
| id | model | now | context | tok/s | in /M | cached /M | out /M |
|---|---|---|---|---|---|---|---|
ds-flash |
DeepSeek V4 Flash | offline | 256k | 400+ | $0.80 | $0.08 | $1.60 |
kimi-k3 |
Kimi K3 | offline | 1M | 460+ | $5.00 | $0.50 | $25.00 |
Default model: ds-flash. Live state on this page is read from the running fleet
at the moment you loaded it — a model with no healthy server is offline, and we say so rather than queue you.
Three ways in
| surface | endpoint | who |
|---|---|---|
| Anthropic Messages | https://api.turboprivate.ai/v1/messages | Claude Code |
| OpenAI Responses | https://api.turboprivate.ai/v1/responses | Codex |
| OpenAI Chat | https://api.turboprivate.ai/v1/chat/completions | every other client |
Same key for all three. Send it as
Authorization: Bearer or
x-api-key — both are accepted on every endpoint, because different clients send different headers.