Rate limits
Rate limits govern how many requests you can make. The following limits apply to every account, regardless of account status.
No international card? Top up with USDT: click Pay with crypto in the top-up dialog and pay with USDT, USDC or ETH from Binance, OKX or any exchange app — the checkout shows which networks your amount supports (BEP20 works from $5). No wallet extension needed (guide).
New accounts start with a $0 balance — top up before your first request. Top up from $5. If a request returns 402 (400 on the OpenAI format) with "wallet balance is insufficient", your balance is empty — your base URL and key are fine. Top up and retry.
Paid models: requests per minute per account
Paid models are limited only by requests per minute (RPM) per account; there is no limit on concurrent connections. Going over the limit returns HTTP 429 (rate_limit_exceeded). The retry-after header is how many seconds to wait, and x-ratelimit-limit is your account's current per-minute limit.
- Occasional 429s: wait for
retry-afterand retry, or lower the number of requests you run at once. - Frequent 429s, or bulk jobs: see the bulk API page — contact sales and we can raise your account's per-minute limit.
Lower-priced -free model limits
If you are using a lower-priced model (with an ID ending in -free; it is still billed), the following limits apply:
Personal daily allowance
| Model ID | Requests per day |
|---|---|
deepseek-flash-free |
200 |
deepseek-v4-flash-free |
200 |
glm-5.3-flash-free |
200 |
Personal allowances reset daily at 00:00 (UTC+8).
Platform-wide daily capacity
Lower-priced -free models are billed at a 99% promotional discount as a limited daily offer: in addition to your personal allowance, the platform provides a limited total capacity for -free models each day, first come, first served.
| Lower-priced lane | Daily capacity |
|---|---|
DeepSeek lane (shared by deepseek-flash-free and deepseek-v4-flash-free) |
~6 billion tokens |
GLM lane (glm-5.3-flash-free) |
~2 billion tokens |
- Your personal allowance and the platform-wide capacity apply at the same time; reaching either pauses access to the
-freemodels for the rest of the day. - During peak hours the daily capacity may run out early. It is not replenished within the same day and resets together with personal allowances at 00:00 (UTC+8), so off-peak usage is recommended.
- Any change to the daily capacity will be announced on this page in advance.
Once a -free model's daily allowance is used up, switch to the corresponding regular model: deepseek-flash, deepseek-v4-flash, or glm-5.3-flash. Regular models are not subject to the daily capacity.