DeepSeek Releases Experimental Multimodal Model on API
DeepSeek states its new experimental model adds vision to the V4-Flash API along with a Files API for image reuse.
TLDR
DeepSeek posted that DeepSeek-V4-Flash-Vision-Exp is now live on its API platform. The model matches V4-Flash on text tasks and shows gains on multimodal agent benchmarks. Users set the model name to deepseek-v4-flash-vision-exp for Chat Completions, Messages, or Responses endpoints. Images count toward billing at up to 384 tokens each. A separate Files API lets developers upload an image once and reference it by file_id across requests at no charge. The company also noted that the model works with existing agent frameworks for combined text and visual tool use.
Combined views
2.7M
21 Sources, first seen ago
Sources
TC 
Tianyi Cui@tianyi可以试试 DSH 最新版啊，内置了新的多模态模型 DeepSeek-V4-Flash-Vision-Exp 支持。 https://twitter.com/deepseek_ai/status/2090730032574631962
- likes: 652
- replies: 76
- bookmarks: 28
- reposts: 28
T( Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex> strictly stronger than 0731 I KNEW IT, IT WAS A BLINDED VLM ALL ALONG https://twitter.com/deepseek_ai/status/2090730032574631962
- likes: 868
- replies: 33
- bookmarks: 83
- reposts: 26
ZP Zizheng Pan@zizhpanToday we announce vision capability for our API model. A milestone, and we will continue to evolve. Please try it and share your experience with us!🎉 https://twitter.com/deepseek_ai/status/2090730032574631962
- likes: 1.8K
- replies: 57
- bookmarks: 112
- reposts: 73
ZE Zephyr@zephyr_z9- likes: 284
- replies: 6
- bookmarks: 11
- reposts: 3
T( Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTexThis is important too Two more notes: - they don’t call models “exp” or “preview” as a joke. 3.2-exp was half a generation behind 3.2. Vision-Full will go crazy - nothing said on open weights so far https://twitter.com/deepseek_ai/status/2090730042586489333
- likes: 319
- replies: 13
- bookmarks: 28
- reposts: 2
CH Chubby♨️@kimmonismuswtf is happening today: DeepSeek has launched DeepSeek-V4-Flash-Vision-Exp, an experimental multimodal model built for agents that need to see. And its performance on visual-agent benchmarks moves close to or even outperforms Opus 4.8. Again: this is the Flash model, the…
- likes: 1.9K
- replies: 65
- bookmarks: 297
- reposts: 129