Help & answers

Models, usage limits, privacy, access and billing.

Models, usage, access and privacy

What exactly is the model?

Two openly-licensed frontier models, both uncensored at the tensor level so refusals are physically removed rather than suppressed: GLM-5.3-Flash, a 328B-parameter sparse flagship with 262K native context, vision, reasoning, and tool calling, and Qwen 3.8 Flash Next, a 125B sparse MoE with only 6B active parameters, so it stays fast at long context, with vision, tool calling and adjustable reasoning intact. Every tier is genuinely unlimited on either model.

Is it really unlimited?

Yes. There are no token caps, usage meters or fair-share throttling. Plans differ only in context window and concurrent streams. Capacity is managed on our side and priced into the flat rate.

What does uncensored mean?

Refusal behaviour is removed from the model weights (tensor-level uncensoring), not bypassed with a jailbreak prompt. Vision, tool calling and reasoning are unaffected.

How fast is it?

Streamed output is faster than reading speed. Prefix caching means continued and repeated conversations start responding almost immediately.

Can I send 262K-token prompts?

Yes. GLM-5.3-Flash serves its full 262,144-token context on every tier, with no lower per-request cap. Qwen 3.8 Flash Next includes 65,536 tokens on Roleplay, 131,072 on Plus, and the full 262,144 on Pro, Ultra and Dedicated.

Which clients does it work with?

OpenAI- and Anthropic-compatible clients, including Claude Code, Codex, OpenCode, Cline, SillyTavern, Hermes, Open WebUI and curl. Most take the base URL https://api.unquant.io/v1 and your API key; Claude Code takes https://api.unquant.io. See the setup guides.

When do I get access?

Invitations are sent in application order as capacity comes online.

Do you log my prompts?

No. We never log your prompts or completions, and we will never build any capability to monitor them. The only thing we record is performance metrics, such as speed, token counts and request timing, so we can keep delivering the best possible experience. We will never store, sell or use your data for any kind of training, ever.

Can I get a refund?

Plans are monthly and can be cancelled at any time. If the service doesn't work for you in the first week, we refund it.