DeepSeek V4: Current Models and PixVerse Workflows
DeepSeek V4 is no longer a forthcoming model. PixVerse’s April 24 review documented the released V4 family. Since then, DeepSeek’s API lineup has evolved again. This update replaces the original launch speculation with the model names and capabilities listed in the official documentation, checked on September 27, 2026.
Current API Models
| API model | Version listed in the documentation | Vision input |
|---|---|---|
deepseek-flash |
DeepSeek-V4.1-Flash | Supported |
deepseek-v4-pro |
DeepSeek-V4-Pro-0813 | Not supported |
Both entries list a 1 million token context window and a maximum output of 384K tokens. The documentation lists thinking and non-thinking modes, JSON output, tool calls, Responses API and Anthropic-compatible API support. Fill-in-the-middle completion is limited to non-thinking mode.
For new Flash integrations, use deepseek-flash. The older deepseek-v4-flash and deepseek-v4-flash-vision-exp names are still accepted, but their original models have been retired: requests are served by V4.1-Flash and billed at the Flash rate.
Vision Does Not Mean Video Generation
Vision input lets a supported model interpret visual material. It does not establish native image or video generation. The current documentation does not support the original article’s claim that one DeepSeek inference generates a picture, video, and accompanying text together. Pro also does not list vision support, so do not transfer Flash capabilities to it.
A long context window can help carry scripts, brand rules, product notes, and revision history. It does not guarantee perfect retrieval or consistency: test the exact documents and tasks your project needs.
Use DeepSeek for Planning, PixVerse for Production
Use the assistant to organize a brief, draft a script, break it into shots, and identify continuity constraints. Review facts and approve the shot plan, then use PixVerse’s image and video models for the visual generation stage. Keep the approved references and outputs together when iterating.
The official PixVerse article discusses tracking DeepSeek for creative workflows. It does not confirm a currently selectable DeepSeek image or video model on PixVerse. Use the model selector and platform announcements to check actual availability.
Check Live Pricing Before Integrating
DeepSeek now distinguishes peak and off-peak prices, cached and uncached input, and output tokens. Check the current pricing and billing rules rather than budgeting from the April launch table. API usage and PixVerse generation are separate stages; this article does not announce shared credits or a PixVerse integration date.