DeepSeek V4-Flash-Vision-Exp Launches: Multimodal API Service Goes Live, Agent Capabilities Approach Opus-4.8

en.Wedoany.com Reported - On August 21, Weidu.com learned from the DeepSeek official website that DeepSeek has just announced the launch of its new multimodal visual understanding model, DeepSeek-V4-Flash-Vision-Exp, on the DeepSeek API platform.

DeepSeek officially stated that V4-Flash-Vision-Exp demonstrates strong multimodal adaptation capabilities across various Agent frameworks, effectively supporting users in leveraging diverse Agent tools to unlock more practical work scenarios.

This is an experimental model, and users can access it by setting model='deepseek-v4-flash-vision-exp'.

Balanced Text and Multimodal Capabilities

According to the introduction, in terms of pure text capabilities (Agent, reasoning, world knowledge, etc.), DeepSeek-V4-Flash-Vision-Exp is on par with the official DeepSeek-V4-Flash release; on Agent Benchmarks requiring visual understanding, DeepSeek-V4-Flash-Vision-Exp achieves a significant leap over DeepSeek-V4-Flash, with multimodal Agent capabilities approaching Opus-4.8.

API Support

Users can access the V4-Flash-Vision-Exp model API by setting model='deepseek-v4-flash-vision-exp'. In the API service, images are converted into tokens and billed accordingly, with each image consuming up to 384 tokens, and the pricing is consistent with the V4-Flash model.

The multimodal API supports three invocation formats: Chat Completions, Messages, and Responses, making it easy to integrate into various Agent tools. It supports mixed text-image input and three image transmission methods: base64 inline, external URLs, and the Files API.

Files API Launch

The Files API is now open and free of charge. Users can upload images to the platform first and then reference them via file_id in requests, thereby saving request bandwidth (the same image does not need to be re-uploaded across multiple requests).