Chinese artificial intelligence company DeepSeek has released an experimental multimodal version of its V4 Flash model that can analyze images, screenshots and visual prompts while preserving its text, reasoning and agent capabilities.
DeepSeek says the model, called V4-Flash-Vision-Exp, performs close to Anthropic's Opus 4.8 on multimodal agentic benchmarks. It is available through DeepSeek's API.
The release comes as the Hangzhou-based startup advances preparations for an initial public offering, with a potential filing as soon as this year. DeepSeek is planning a mainland China IPO, with a filing targeted for 2026 that could allow the company to debut publicly in 2027, Bloomberg reported.
The company has started discussions with accounting firms and investment banks as it advances its plans.
DeepSeek is also seeking additional private funding ahead of a potential listing, shortly after completing a reported $7 billion round. The company is in talks with investors for a new round targeting a pre-money valuation of at least 480 billion yuan, or about $71 billion, up from roughly $50 billion in its first external round, which included Tencent Holdings (HKG:0700, OTC:TCEHY) and Contemporary Amperex Technology, Bloomberg reported.