minimax-h3 · MiniMax · video
Official MiniMax model; specifications and prices verified against official documentation on 2026-10-07. Asynchronous video generation.
Input / output / endpoint: `text, image, video, audio` → `video`. Native V2 asynchronous tasks; content accepts text/image/video/audio; task_id → GET /v2/query/video_generation; status and output assets must be validated; 768P / 2K. duration: 4..15 seconds; request body ≤64MB; reference images ≤9 (30MB each); videos ≤3, total ≤15 seconds (50MB each); audios ≤3, total ≤15 seconds (15MB each).
Short video ads, product demonstrations, story shots, and image animation; supported reference inputs depend on the model.
The key must have access to the model. Included subscription usage follows the MiniMax console. Official sources: [https://platform.minimax.cn/docs/api-reference/video-generation-v2-create](https://platform.minimax.cn/docs/api-reference/video-generation-v2-create) · [https://platform.minimax.cn/docs/guides/pricing-paygo](https://platform.minimax.cn/docs/guides/pricing-paygo) · [https://platform.minimax.io/docs/guides/pricing-paygo](https://platform.minimax.io/docs/guides/pricing-paygo) · [https://platform.minimax.cn/docs/guides/pricing-token-plan](https://platform.minimax.cn/docs/guides/pricing-token-plan) Native forwarding is implemented. Images support URL/base64 responses, one character reference and JSON edits; multipart edits, masks and image streaming are unavailable. Video is billed on successful output. H3 accepts up to 5 reference images; H3-Max up to 2. Separately charged video references and excess images are rejected before dispatch. Regeneration and Context-IR remain catalog-only.
Anthropic official CLI coding agent. Multi-file edits, tool use, long-context reasoning.
export ANTHROPIC_AUTH_TOKEN="YOUR_API_KEY"
export ANTHROPIC_BASE_URL="https://api.robovai.com"
export ANTHROPIC_MODEL="minimax-h3"
claude