Models API

Updated

GET /v1/models

OpenAI format, key required. Returns only models this key may call (subject to the key's allowed models) that have an available endpoint.

curl https://modelportal.ai/v1/models -H "Authorization: Bearer YOUR_API_KEY"

{"object": "list", "data": [{"id": "deepseek/deepseek-v3.2", "object": "model", "created": 1767225600, "owned_by": "modelportal"}]}

The Anthropic format (with an anthropic-version header) and the Gemini format (GET /v1beta/models) follow the same rules.

GET /api/v1/models

Platform format, no auth needed.

Query parameter Meaning
q Search by model name, model ID or publisher
capability Filter by capability: tools, vision, reasoning, structured_output
provider Filter by provider identifier
data_policy, data_region data_policy takes no_retention or no_training; data_region takes us, eu, apac or cn
limit, cursor Pagination; when the response has next_cursor, pass it to get the next page

Each model carries slug (the model ID), display_name, capabilities, status, availability, input_limits, price_summary, provider_count, publisher details and more, plus offerings: one entry per routable endpoint with the fields listed under the detail below, and also:

Field Meaning
config_version The current configuration version; the current price version is price.version
rpm_limit, tpm_limit Declared rate limits; null when not declared
stats_24h Success rate and time to first token over 24 hours; null with too few samples
rating Eagle Eye rating summary; not connected yet, so null

Responses may be cached for up to 60 seconds.

GET /api/v1/models/{model ID}

One model in detail, no auth needed; write the / in the model ID as %2F. Besides the list fields, it returns offerings, one entry per routable endpoint:

Field Meaning
offering_id Endpoint identifier, starting with route_; usable in provider.only
supplier.slug Provider identifier; usable in provider.order, only and ignore
status Endpoint status, such as available or degraded
variant, upstream_model, source_type Variant name, upstream model name, supply type
config Declared configuration: quantization, context_length, max_output_tokens, capabilities, sampling_params and more
data_policy Retention, retention days, training use, data processing region and privacy policy link; undeclared items are null
forbidden_regions Regions the endpoint does not serve
price, next_price The price version in effect and the earliest pending one
stats_7d Success rate and time to first token over 7 days; null with too few samples 
curl https://modelportal.ai/api/v1/models/deepseek%2Fdeepseek-v3.2