mimile
Back to feed

DeepSeek releases V4-Flash-Vision-Exp, an experimental multimodal model

AI digest

This digest was compiled by AI from multiple sources — links to the originals are below.

DeepSeek releases V4-Flash-Vision-Exp, an experimental multimodal model

Chinese AI company DeepSeek has released V4-Flash-Vision-Exp, an experimental multimodal model that adds image understanding to its text capabilities. On DeepSeek's internal benchmarks, the model nearly matches Opus 4.8 on agent tasks. The release targets visual agent workflows, with support for OpenAI and Anthropic APIs and a new free Files API.

Key Facts

  • DeepSeek released V4-Flash-Vision-Exp, an experimental multimodal model extending V4-Flash with image processing.
  • On DeepSeek's internal multimodal agent benchmarks, the vision variant scores close to Opus 4.8.
  • The model supports JPEG, PNG, GIF, and WebP formats, and determines format from file content rather than filename or MIME type.
  • A single request can include up to 600 images, with a maximum edge length of 8,192 pixels per side, dropping to 4,096 pixels for requests with 15 or more images.
  • Each image costs at most 384 tokens, and pricing follows V4-Flash rates.

Model Capabilities

DeepSeek-V4-Flash-Vision-Exp extends DeepSeek-V4-Flash with image processing while keeping the base model's text performance in reasoning and world knowledge, according to DeepSeek. The model is designed to work with different agent frameworks and combine visual understanding with tool use. In practice, it can describe images, extract text from screenshots, and analyze diagrams. It handles JPEG, PNG, GIF, and WebP, and determines the format from actual file content rather than the filename or declared MIME type, per the API docs.

API Integration and Pricing

The model works with OpenAI's Chat Completions and Responses APIs and Anthropic's Messages endpoint. DeepSeek also released version 0.1.1 of its Harness framework, which supports the new model out of the box. Developers can send images via Base64 encoding, publicly accessible URLs up to 32 MiB, or the new free Files API with a 64 MiB size limit. An optional detail field downscales images to 512 x 512 pixels, saving tokens when fine visual detail is not needed. The model automatically normalizes images to roughly 800 x 800 pixels depending on aspect ratio before processing, and each image costs at most 384 tokens.

1 source

DeepSeek releases V4-Flash-Vision-Exp, an experimental multimodal model

the-decoder.com

The Decoder