
We added 12 models in October: two text, seven image and three video. You can choose new text models or use reference media to generate and edit images and videos.
Try the new text models
Claude Haiku 5.5 supports image input, function calling and the native Messages API. Gemini 4 Argon accepts images, video, audio and PDFs, with a native Gemini API option. See the text model documentation for parameters and limits.
Generate and edit images with references
New options include FLUX 3 Image and Klein, Nano Banana 2.1 in official and economy versions, MAI-Image-2.6 and Flash, and HY Image 3.5 Preview.
All support image editing, with different reference limits, output resolutions and billing rules. For example, Nano Banana 2.1's official version is billed by actual token usage, while its economy version uses a fixed per-image rate. Check the model page before choosing a route.
Generate, edit or extend video
Kling 4.0 supports generation, editing and extension at 720p or 1080p for clips of 3–30 seconds, always with audio. We also added Vidu Q4 Preview, which accepts image and audio references with output up to 4K, and PixVerse V6 Max for text-to-video and single-image animation.

Refresh your cost estimates
- October 10: Seedance 2.0 Mini and Fast rates increased.
- October 18: 2K and 4K rates increased on selected GPT Image economy models.
- October 28: several Nano Banana economy rates changed. Official versions billed by token usage were outside this adjustment.
See the model changelog and pricing changelog for the full record. Open the model catalog to choose a model and check its parameters and quote before making a call.


