Skip to main content
Chat Completions for GPT-5, Grok, Gemini, and DeepSeek; Responses for GPT-5 and Grok; Gemini native generateContent for Gemini. Each model page lists its protocols. FreeModel does not translate between protocols.
Not yet. Claude is listed as coming soon; when it launches it will be served on Anthropic Messages (POST /v1/messages).
When a GPT-5 request’s prompt (input plus cached tokens) exceeds 272,000 tokens, the whole request bills at that model’s long-context rate, listed on each GPT-5 model page.
Seedance 2.0 comes in standard (480p to 4K), Fast, and Mini lines (480p or 720p), each with t2v, i2v, and r2v models and 4 to 15 second output. Seedance 2.5 offers 480p and 720p SKUs (with -v2v edit SKUs) and t2v, i2v, r2v, and extend mode models with 4 to 30 second output.
Gemini Omni bills per fixed 10-second request at 720p or 1080p and uses a pixel size. Flash 1.1 bills per second, takes 4, 6, 8, or 10 seconds, offers 360p to 4K through resolution, uses aspect_ratio, and accepts up to seven reference images.
Model IDs without -4k produce 1080p. veo3-1-lite-4k, veo3-1-fast-4k, and veo3-1-quality-4k produce 4K. All Veo routes are fixed at 8 seconds. Lite, Fast, and Quality accept zero to two images; veo3-1-fast-ref requires one to three.
gpt-image-2 (quality low, medium, or high and a pixel size) and GPT Image 2.5 Ext (gpt-image-2.5-ext and gpt-image-2.5-ext-sunburst, with 1K, 2K, or 4K output, ratios, up to four images, and up to 16 references).
No. Image, video, and audio inputs must be public http or https URLs; private addresses and local files are rejected.
Call GET /v1/models at runtime and check the live models page for current prices.