Docs
Model list
Read the catalog from code. Pricing, context and modalities come with it, so use this when building your own model picker.
Listing
GET /models returns everything. Models slated for sunset are excluded by default; add include_deprecated=true to include them.
Response fields
Prices are USD per million tokens. Billing converts to won at the exchange rate in effect when the call is made.
| Field | Type | Note |
|---|---|---|
| id | string | Request model identifier |
| provider | string | Owning provider |
| context_window | integer | Maximum input tokens |
| max_output_tokens | integer | Maximum output tokens |
| input_price | number | USD per 1M input tokens |
| output_price | number | USD per 1M output tokens |
| modalities | array | text, vision, audio, image |
| region | string | Inference region |
Filters
Narrow by provider and modality. Models on your organization denylist never appear at all.
Cache and long-context prices
pricing carries cache_write_1h, the 1-hour cache write price, next to cache_write, the 5-minute one. Models with a long-context tier return long_context with the token threshold (threshold_tokens) and the long-context prices; when the prompt tokens (input + cache read + cache write) exceed the threshold, the whole request is billed at those prices. long_context is null for models without the tier.