DeepSeek ships experimental vision model and Files API
- Source
- x.com
- Date
DeepSeek-V4-Flash-Vision-Exp is now live on the DeepSeek API Platform! 🚀 🔹 This experimental multimodal model matches DeepSeek-V4-Flash on text capabilities—including agents, reasoning, and world knowledge. 🔹 On multimodal agent benchmarks, V4-Flash-Vision-Exp makes a major leap over V4-Flash, bringing multimodal agent performance close to Opus-4.8. Try it with model='deepseek-v4-flash-vision-exp'. DeepSeek Harness 0.1.1 was released today with out-of-the-box support for the new model. 1/n

Context
DeepSeek has added image input to its API through an experimental model, deepseek-v4-flash-vision-exp, which accepts text and images in the same request. In its release note, DeepSeek says the model matches the text-only DeepSeek-V4-Flash on agent, reasoning and world-knowledge tasks. It also says that on agent benchmarks the model comes close to Opus-4.8 from Anthropic. The note names no benchmarks and gives no scores in its text, so that comparison is DeepSeek's own claim and could not be checked here.
The new Files API matters most for repeated work. Images can be sent as base64-encoded data inside the request, as an external URL, or as a reference to a already uploaded. With the Files API, an image is uploaded once and then cited by its file ID in later requests, which DeepSeek says saves request bandwidth. The Files API itself is free, but images still cost tokens: DeepSeek bills each image at up to 384 tokens at V4-Flash prices. Reusing an uploaded file reduces what is sent with each request, not what is charged for it.
The model is available through the Chat Completions, Messages and Responses endpoints, and DeepSeek 0.1.1, released the same day, supports it out of the box. The experimental label is the main caution for anyone building on it, because the release note sets no timeline for a stable version or for retirement.
- multimodal — A model that works with more than text — reading images, audio, or video, and sometimes generating them too.
- File — A named unit of stored data that keeps its contents while no program is using it.
- Harness — The software that sends a model its context, recognizes tool calls, runs or refuses them, returns results and keeps the record.
- Multimodality — The ability to understand or generate more than one kind of media.
Checking sign-in…
Loading comments…





