Skip to main content
OverControl prices text requests by the model and by the tokens in the call. Prices below are per million tokens. Cached input is a portion of the input, billed at the lower cached rate. Tokens spent creating a cache are billed at the ordinary input rate. There is no separate cache-write price.
Included text requests scale with the plan. Plus is 5× the Free allowance, and Pro is 20×. Chat Completions and Messages share that allowance for the UTC calendar month. Listing models, and calls to the Files API, do not spend it.

​
Model pricing

You are billed in euros. Context windows and output caps are on Models. Lume 3 matches Lume 3.5 on context, output, and price. Lume 3 Max matches Lume 3.5 Max.

​
How a request is priced

The cost of a request comes from three kinds of tokens.
  • Input tokens are what you send: instructions, messages, and the rest of the request.
  • Cached input tokens are the part of that input served at the model’s lower cached rate. They are already inside the input count.
  • Output tokens are the tokens the model writes in its reply.
On Chat Completions, usage.prompt_tokens is the full input. usage.prompt_tokens_details.cached_tokens appears when the cached count is greater than zero. Suppose a Lume 3.5 request uses 100,000 input tokens, none of them cached, and 2,000 output tokens:
If 50,000 of those input tokens are cached, the input charge is 50,000 at €0.30 and 50,000 at €0.06, which is €0.018. Output stays €0.003. The request comes to €0.021. On Messages, input_tokens and cache_creation_input_tokens are charged at the input rate. cache_read_input_tokens is charged at the cached rate. Output uses output_tokens. The euro rates are the same ones in the table. Caching can show up on a response, and there is no request field that turns it on. cache_control does not create a cache.
For a busy application, the model and any cached input change the bill more than small differences in prompt wording.

​
Included requests

Chat Completions and Messages draw from one allowance each UTC calendar month. How much you get depends on the plan, as a multiple of Free. Once that allowance is used, further text requests are billed in euros at the rates above. If a request cannot be billed, you’ll get HTTP 429. OverControl does not publish a requests-per-minute cap, and these responses do not carry x-ratelimit headers. On Chat Completions the body looks like this. error.type is rate_limit_error.
Messages uses the same words, in its own shape. error.type is rate_limit_error, and there is no error.code.
Both responses include X-Request-Id. Messages also sends request-id. Usage for the current month is in the API dashboard.

​
Next

Models

Compare the Lume models and pick one for the workload.

Quickstart

Create a key and make a first request.