Quick Summary
- DeepSeek unveiled two new open-source AI models: V4-Pro (1.6 trillion parameters) and V4-Flash (284 billion parameters)
- Each model features a one million token context window, comparable to Google’s Gemini capabilities
- V4-Pro achieves performance parity with OpenAI’s GPT-5.4 in coding tests and ranks second only to Gemini in reasoning tasks
- The company emphasizes significantly reduced computational and memory requirements versus competing models
- Launch timing coincides with reports of Tencent and Alibaba pursuing investment at over $20 billion valuation
Chinese artificial intelligence company DeepSeek unveiled preview editions of its latest flagship open-source model, V4, this past Friday. According to the firm, the new release delivers enhanced reasoning capabilities, reduced operational costs, and an exceptionally large context window.
DeepSeek introduced two distinct variants: V4-Pro and V4-Flash. The Pro variant features 1.6 trillion parameters, while the Flash variant represents a streamlined alternative with 284 billion parameters, optimized for efficiency and cost-effectiveness.
Each variant supports a context window capable of handling one million tokens. This capability enables the models to analyze substantial volumes of text simultaneously, positioning them competitively alongside Google’s Gemini in this dimension.
DeepSeek indicated the current models operate exclusively with text input. The organization noted ongoing development to incorporate multimodal functionality, enabling future processing of visual content including images and video.
Performance Against Competitors
In MMLU-Pro testing, a prominent AI evaluation benchmark, V4-Pro delivered results equivalent to OpenAI’s GPT-5.4. The model scored marginally below Google’s Gemini and Anthropic’s Claude Opus 4.6. For reasoning-focused benchmarks, V4-Pro secured second place, trailing only the most recent Gemini release.
DeepSeek further noted V4’s optimization for AI agent frameworks, including Claude Code, OpenCode, and CodeBuddy compatibility.
The organization characterized V4’s context capabilities as “world leading with drastically reduced compute and memory costs.” Industry analyst Zhang Yi described the development as an “inflection point,” suggesting ultra-long context functionality could transition from academic research into mainstream commercial applications.
AI industry analyst Max Liu characterized the launch as a “milestone” for China’s artificial intelligence sector, drawing parallels to the market impact when DeepSeek’s R1 initially debuted.
Industry and Financial Implications
This marks DeepSeek’s first comprehensive new model architecture since R1’s introduction in early 2025. That previous release disrupted global technology equities, including Nvidia and Meta, by demonstrating that cost-efficient, streamlined models could effectively compete with premium closed-source alternatives.
DeepSeek has not disclosed which semiconductor chips powered V4’s training process. Earlier in the current year, U.S. authorities alleged the company utilized prohibited Nvidia Blackwell processors. Subsequent reporting from The Information suggested the models underwent training using Huawei chips instead.
Huawei verified that its Ascend supernode infrastructure, powered by Ascend 950 AI processors, would provide comprehensive support for DeepSeek’s V4 model family.
The model debut follows recent reports indicating Tencent and Alibaba have entered negotiations to acquire stakes in DeepSeek at a valuation exceeding $20 billion. DeepSeek ranks among China’s six premier AI unicorn companies.
A preview build of V4 is currently accessible through Hugging Face. DeepSeek has yet to specify a timeline for the complete public release.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
