How it works.

Choose the enabled abliterated model and a funded duration from 15 minutes to 24 hours. Prepaid wallet credit is reserved before Start. There’s no hardware to buy or model server to configure.

Every user gets a dedicated GPU instance. Your instance is not shared with other customers. We handle model setup and serving.

Your model stays fixed for the session. Use your own client or harness for external tools and web access; the model API does not run those tools for you.

Before release, each model is tested for refusal behavior, coding and tool use through the same API, quantization and serving configuration you use. Our service adds no content filters or refusal instructions.

We handle compute and model setup. Startup usually takes about 5 minutes and can take longer. Paid time begins only at READY. Fund at least 15 minutes; extend explicitly if you need more time.

Billing uses elapsed READY seconds, including idle time; startup is not billed. End session or funded expiry stops access and billing. Tokens don’t add to the bill.

Unused reserved credit returns to your wallet when the session ends. A detected service failure stops billing at its recorded failure time and returns unused reserved credit.

During the invite-only preview, credit is added manually. Checkout, deposits and bonuses are not offered. Extend while READY to add funded time with wallet credit.

See hourly pricing

Start a session and wait until it is ready. Connect through an OpenAI-compatible client or your own harness. Your API endpoint and account key stay the same across sessions. Native tool calls are returned for your client to execute; no hosted agent, shell, browser, files or MCP server is included.

There are no per-token charges. Qwen 27B context and a measured decode result are listed on the pricing page. Qwen Flash-Next and GLM Flash are not available yet.

Run one generation at a time per session. Parallel generation requests return 429 GENERATION_BUSY; wait for the active request to finish before retrying.

Human technical support is planned for launch. Live chat is not connected in this preview.

Save the prompts and responses you need in your own client. There is no hosted conversation history, workspace or export service to reopen after the session ends.

Closing the browser or canceling a request does not end the session. Use End session to stop API access and ready-time billing. Funded time begins only at READY and stops at End, expiry or a detected runtime failure.

Your client controls any saved conversations, files and external tool results. Those copies are outside session cleanup.

The dashboard shows the funded deadline. Save needed responses before expiry, or explicitly extend with wallet credit.

What session privacy covers