DeepSeek shipped deepseek-v4-flash-vision-exp on its API platform. While matching V4-Flash on text reasoning and coding, its multimodal agent scores jump close to Claude Opus 4.8. Images bill at standard V4-Flash rates (up to 384 tokens per image) with day-one support for free Files API and DeepSeek Harness 0.1.1.

Key Takeaways

  • ✓Exposes `deepseek-v4-flash-vision-exp` endpoint with multimodal agent accuracy near Opus 4.8.
  • ✓Billed at ultra-low V4-Flash rates (max 384 tokens per image) with free Files API.
  • ✓DeepSeek Harness 0.1.1 brings unified support for Chat Completions and Responses APIs.
ADSponsored