LU Cloud product facts

Flash chat: published daily allowance

Hosted includes 11 Flash models without credits; Pro and Max include 12. Unlimited prompt counts, up to 500,000 combined input and output tokens per account per UTC day in app sessions. All 48 chat models remain available on every plan. GLM 5.3 Flash uses credits on Hosted and is included in the daily allowance on Pro and Max. API keys and accounts without an active subscription always use credits.

App sessions share 500,000 input and output tokens per account per UTC day across the Flash models included on your plan without spending credits. The allowance resets at 00:00 UTC. It is a benefit of an active paid plan, not a free tier: an account without an active plan uses credits for these models, whatever it bought before. API key requests always use credits.

One free request can run at a time per account. A second concurrent request is refused, not silently billed. A request whose token budget exceeds the remaining allowance uses normal credit pricing, with a notice in the app.

We reserve the input estimate and output limit before generation. Without an explicit output limit, free requests allow up to 8,192 output tokens. Each free request lasts at most 240 seconds. Final provider usage settles the reservation. Failed or interrupted requests without final usage keep their reservation. This is a daily allowance, not unlimited usage.

Request limits, not hidden windows

• Chat: 60 requests per minute per account.
• Media job submissions: 30 requests per minute per account.
• Media uploads: 60 requests per minute per account.
• Subscription checkout: 5 attempts per hour and 12 per day per account.
• Plan changes: price previews are not counted against that; payment attempts are limited to 6 per hour and 12 per day per account, and all checkout requests together to 40 per hour.
• Credit-pack checkout: a separate 5 attempts per hour and 12 per day per account.

These are request counters, not guaranteed completed generations. They use windows starting with the first request and count requests that reach the limiter, including later validation failures. A rate-limit response includes a Retry-After header.

These burst counters currently live in each server process. Restarts reset them, and multiple server instances do not share them. They are not a global concurrency or availability guarantee. Provider capacity and wallet checks can also refuse a request. The separate Flash daily allowance and one-request limit above are enforced in the database across instances.

Chat model capabilities

Every model on every plan. All 48 chat models are available on paid plans and credit packs. Context and quantization below are what the model host lists, checked 2026-09-11. Missing quantization is not a claim of full precision.

Tool transport, image input and reasoning controls describe the LU catalog, not a new live capability test. Provider-listed context is not a guarantee of a full-length prompt: input and output share context, and clients may trim history. No full-native-context promise is made here.