Geo News
Community curated by people like you
LatestAICryptoHealthWorld AffairsUS Politics
DeepSeek Previews V4 Model Series with Million-Token Context Windows
00

DeepSeek Previews V4 Model Series with Million-Token Context Windows

Apr 24, 2026

Chinese AI startup DeepSeek released preview versions of its V4-Pro and V4-Flash models featuring 1.6 trillion parameters and one-million-token context windows, claiming to close the gap with frontier models while maintaining significantly lower pricing than US competitors. The V4-Pro model significantly leads other open source models in world knowledge benchmarks and is only slightly outperformed by Google's Gemini-3.1-Pro. The release comes amid escalating tensions, with the White House accusing China of industrial-scale IP theft and OpenAI, Anthropic, and Google warning of model extraction attacks by Chinese firms including DeepSeek. Multiple Western governments including the US, Italy, South Korea, and Germany have banned or restricted DeepSeek citing national security concerns and illegal data transfers to China.

DeepSeek V4 Model Capabilities and Performance

  • ▪DeepSeek V4 can be used with AI agents including Claude Code, OpenClaw and OpenCode
  • ▪DeepSeek released the V4 model as DeepSeek V4-Pro and DeepSeek V4-Flash

Global AI Competition and Strategic Positioning

  • ▪DeepSeek gained attention in late 2024 with its open source V3 model
  • ▪DeepSeek released R1 in January 2025, an AI reasoning model that DeepSeek claimed was cheaper and performed just as well as OpenAI's rival ChatGPT
  • ▪OpenAI launched GPT-5.5 one day before DeepSeek released V4
  • ▪DeepSeek released models in January 2025 that appeared to deliver strong performance at lower cost than many US rivals
  • ▪OpenAI, Anthropic, and Google warned in 2025 that Chinese AI firms including DeepSeek were conducting model extraction attacks

Security Concerns and Geopolitical Restrictions

  • ▪Germany banned DeepSeek in Apple and Google app stores in 2025 citing illegal transfer of user data to China
  • ▪The United States banned government agencies from using DeepSeek in 2024 citing national security concerns

Perspective of US AI companies and government officials

  • ▪OpenAI, Anthropic, and Google warned in 2025 that Chinese AI firms including DeepSeek were conducting model extraction attacks

Perspective of Western government regulators

  • ▪Germany banned DeepSeek in Apple and Google app stores in 2025 citing illegal transfer of user data to China
  • ▪The United States banned government agencies from using DeepSeek in 2024 citing national security concerns

8 sources

Techcrunch
DeepSeek previews new AI model that ‘closes the gap’ with frontier models
View source article
Mezha
DeepSeek launches V4 Flash and V4 Pro, challenges top AI models | Ukraine news - #Mezha
View source article
The-decoder
As agentic AI pushes rivals to raise prices and cap usage, Deepseek ships a good-enough model for almost nothing
View source article
Newsbytesapp
China's DeepSeek previews new AI model to rival US giants
View source article
Letsdatascience
DeepSeek previews V4 Pro and V4 Flash models | Let's Data Science
View source article

Featured stories

China Blocks Meta's $2 Billion Manus Acquisition, Bars Founders from Leaving Country

Apr 17, 2026 · 1 source

White House Accuses China of Industrial-Scale AI Model Theft Campaign

Apr 24, 2026 · 2 sources

DeepSeek Seeks $300 Million in First Fundraise at $10+ Billion Valuation

Apr 18, 2026 · 1 source

Story comments

Loading comments…

Related entities

United StatesChina

Related Projects

DeepSeek

Topics

China AI regulationsAI startupsAI antitrust & competitionAI foundation modelsLarge language models (LLMs)Long context models

Featured stories

China Blocks Meta's $2 Billion Manus Acquisition, Bars Founders from Leaving Country

Apr 17, 2026 · 1 source

White House Accuses China of Industrial-Scale AI Model Theft Campaign

Apr 24, 2026 · 2 sources

DeepSeek Seeks $300 Million in First Fundraise at $10+ Billion Valuation

Apr 18, 2026 · 1 source