SDK 6.29.0 catalog update
This update compares SDK 6.29.0 with the 6.23.0 catalog published here on October 2, 2026. There are 3 additions and 4 removals, bringing the production catalog to 222 models across 29 providers: 72 image, 88 video, 31 audio, 30 text, and 1 JSON model.
New production models
| Model ID | Name | Mode | Provider |
|---|---|---|---|
gemini-nano-banana-2.1 | Nano Banana 2.1 | image | |
heygen-video-1 | HeyGen Video | video | HeyGen |
typesafe-evaluate | Jev | json | TypeSafe |
Nano Banana 2.1 supports image generation and editing, up to 14 reference images, 1K/2K/4K resolutions, and a configurable thinking level. HeyGen Video supports a first frame plus image, video, and audio references, 5–15 second durations, and 480p or 768p output.
Jev adds the JSON output mode for structured content evaluation. Use the SDK's ai.run() API; generate() and generateText() do not accept JSON models. See JSON & evaluation.
Removed from the production catalog
luma-ray-2— Luma Ray 2luma-ray-2-reframe-video— Luma Ray 2 Reframeluma-ray-flash-2— Luma Flash 2luma-ray-flash-2-reframe-video— Luma Flash 2 Reframe
The Luma reference now uses the remaining Ray 3.2 family. Its current generation settings are 540p/720p/1080p and 5 or 10 seconds; old Ray 2 settings should not be copied unchanged.
Version and availability
npm install @picsart/ai-sdk@6.29.0This is a production SDK catalog snapshot. Preview, disabled, and deprecated entries are excluded. CLI, MCP, and Playground availability depend on their installed or deployed versions; catalog presence does not guarantee support on every surface.
Check gen-ai models and gen-ai models info <id> --json for CLI support, or the connected MCP server's model discovery tools. For JSON models, use the execution method documented by that interface.