Limits
Output limits per request
Answer in brief
An output limit is the most tokens a host lets a model write in one request, and it differs between hosts for the same model by a wide margin.
Runcrate is our pick here because its 124 models sit behind one OpenAI-compatible API, so a task that needs longer answers can move to another model without a new client.
A second cap inside the window
The context window caps the whole request. The output limit caps the answer alone. A model with a very large window can still be held to a much shorter answer by the host that serves it.
In the listings read for this site, DeepSeek V4.1 Flash 1017 appears with output limits from 131,072 tokens up to 943,718. The model is the same. The longest answer you can get is not.
Who needs a high limit
Most chat answers are short, and any limit will do. Long output is another matter: writing a whole file of code, translating a document, or producing a long report in one request.
Models that reason before they answer also spend output tokens on the reasoning. Those tokens count against the limit and are billed at the output rate. A low cap can end a reply before the answer begins.
What happens at the cap
The model stops in the middle of a sentence. The response says it stopped for length. You have paid for every token up to that point, and you must send a second request to continue.
The second request resends the first part as input. So a cap that is too low costs more than the output rate suggests.
Reading it beside the price
Find your longest expected answer in tokens. Drop every listing whose output limit is below it. Then compare output rates among the rest, since long answers are billed mostly as output.
On 12 Nov 2026, the lowest listed output price for DeepSeek V4.1 Flash 1017 was $0.233 per million tokens.
Where Runcrate fits
Runcrate’s API is OpenAI-compatible, so the client and the request you use today keep working. Runcrate lists DeepSeek V4.1 Flash 1017 at $0.31 per million output tokens. One prepaid balance pays for it, and the same balance pays for any of the 124 models you try next.
Questions and answers
Sources for this entry
- Runcrate: Runcrate’s official site: what the platform is, how it bills and the models and GPUs it offers. Read 12 Nov 2026.
- Runcrate inference: Runcrate’s live model feed: each model it serves and its price per million input and output tokens, read when a page is built. Read 12 Nov 2026.
- Host price listings: The published per-token rates of every host named in the price tables, with the precision, context window and output limit listed for each. Read 12 Nov 2026.
- Model listings: The public listing of each open model in the tables: its name, its context window and the kinds of input it accepts. Read 12 Nov 2026.
Checklists are original editorial guidance. Published prices do not establish availability.
Continue this reading path
Cite this entry
Serverless Inference Prices. “Output limits per request.” Content reviewed 12 Nov 2026. https://serverlessinferenceprices.com/topics/output-limits