Every model has exactly one name.
The name is made up of provider and model. You call the model with it, and under it you find price, context and capabilities in GET /v1/models.
anthropic
Provider
claude-fable-5-1
Model at the provider
This ID goes in the model field. If the catalog does not know it, the API answers with 404 model_not_found.
The catalog lists only what you can call.
A model only appears once it is released and has a price. Every model belongs to one kind: chat, embedding, audio, video or image. If you ask an endpoint for the wrong kind of model, you get 400 unsupported_for_model.
The list as an API.
You can read the list from your code and, for example, check at startup whether a model is still offered. The list comes in pages with limit and after, at most 100 entries per page.
id- The model ID in the provider/model format
pricing- Input and output in euros per 1M tokens, for video per second, for images per image
kind- chat, embedding, audio, video or image
context_length- Input window as stated by the provider
flowai_limits- What FlowModels actually accepts per request
capabilities- Capabilities as individual values
supported_parameters- Fields this model accepts
deprecated_at- Deprecation cutoff date, or null
{
"object": "list",
"data": [
{
"id": "anthropic/claude-fable-5-1",
"object": "model",
"created": 1735689600,
"owned_by": "creativeskyline",
"pricing": {
"prompt": "<euros per 1M input tokens>",
"completion": "<euros per 1M output tokens>",
"currency": "EUR",
"unit": "per_million_tokens"
},
"kind": "chat",
"context_length": 128000,
"max_output_tokens": 16000,
"capabilities": {
"tools": true,
"vision": true,
"reasoning": false,
"web_search": false,
"json_schema": true,
"streaming": true
},
"supported_parameters": ["temperature", "max_tokens", "tools"],
"flowai_limits": {
"input_tokens": 112000,
"max_output_tokens": 16000
},
"provider_metadata": {},
"deprecated_at": null
}
],
"has_more": false,
"first_id": "anthropic/claude-fable-5-1",
"last_id": "anthropic/claude-fable-5-1",
"flowai_request_id": "01HXYZ8M3F6V2KQ"
}
Capabilities are listed per model.
A field a model cannot handle is not quietly ignored: the API rejects it with 400 unsupported_for_model.
| Value in capabilities | The model can |
|---|---|
tools |
call your own tools |
tool_choice |
have its tool choice controlled |
parallel_tool_calls |
call several tools in one response |
vision |
read images in the input |
json_object |
return JSON |
json_schema |
return JSON that follows a fixed schema |
reasoning |
work with adjustable compute effort |
web_search |
search the web |
streaming |
send the response piece by piece |
A model does not disappear overnight.
Models are retired by their providers. FlowModels announces this before a call fails and gives you the cutoff date.
-
Listed
The model is in
GET /v1/modelsand can be called. -
Deprecated
Every response carries the
Deprecation: trueandSunsetheaders with the cutoff date. The call runs normally. -
After the cutoff date
The call is rejected with
410 model_deprecatedbefore credit is reserved or a provider is called.