Models, limits & errors
Model IDs, prices, capabilities, limits, error messages and the features of the API.
Generate and edit images: Playground, API, models and costs
Generate images from text or edit a template: in the Playground or via the API, models, formats and quality, download, seven days of retention and billing per image.
Generate videos: Playground, API, models and costs
Generate videos from text or a start image: in the Playground or via the API, models and parameters, status, download, seven days of retention and billing per second.
History too long: context_length_exceeded and history compaction
What context_length_exceeded means and how to solve it: shorten the history, allow context_compression or let the Brain manage the history.
Create embeddings: turn text into vectors
Turn text into vectors with POST /embeddings: available models, batches of up to 2,048 entries, base64 format and costs in euros.
Let the model call your own functions: tools and tool_calls
Function calling with FlowModels: offer functions, evaluate tool_calls, send results back and avoid common mistakes.
Get responses as JSON: response_format with json_object and json_schema
With response_format you get valid JSON or JSON that follows your schema. How to use json_object and json_schema and how to tell which models support them.
Live transcription: audio in real time over WebSocket
Live sessions transcribe running audio over WebSocket. They are available to you when your model list includes a live model.
What a model can do: check capabilities, limits, output length and reasoning effort
What the model list tells you about a model: capabilities, context, limits, max_tokens and reasoning_effort. How to avoid unsupported_for_model.
FlowModels models and prices
Which models are enabled, how to specify the model ID in a call, how the euro prices per million tokens, per second of video or per image come about, and what happens with retired models.
Request limit: how the tiers work
The four FlowModels tiers with requests per minute and concurrent requests, how your team moves up automatically, and what to do about HTTP 429.
Retry requests safely: Idempotency-Key, Retry-After and rate limit headers
With Idempotency-Key, Retry-After and the rate limit headers you can retry requests without paying twice. Includes sample code in Python.
Send images, PDFs and audio to a model
How to send images, PDFs, Office files and audio directly to a model: via URL, Base64 or a previously uploaded file, with all size limits.
Transcribe audio files: models, sizes and background jobs
Turn audio files into text: models, synchronous and background jobs up to eight hours, timestamps, costs and retention.
Understanding FlowModels error messages
How an error response is structured and the most common codes: 429 for limits and budget, 503 for missing capacity, 400 for malformed requests, plus 401, 403, 404 and 410.
Understanding model IDs: the provider/model format
Every model in FlowModels has a full identifier made up of provider and model. How to read it, find it and use it correctly in the model field.
Turn on web search, limit it to domains and read the sources
How to turn on the provider's web search, limit it to specific domains with search_domain_filter, and read citations and search_results.