DeepSeek previews new AI model that ‘closes the gap’ with frontier models
TechCrunch·2h ago·Hope
What Happened
Chinese AI lab DeepSeek unveiled two preview versions of its DeepSeek V4 large language model—V4 Flash and V4 Pro. Both feature 1-million-token context windows and mixture-of-experts architecture. V4 Pro contains 1.6 trillion parameters (49 billion active), making it the largest open-weight model. DeepSeek claims both models nearly match frontier AI systems on reasoning benchmarks and outperform OpenAI GPT-5.2 and Gemini 3.0 Pro on specific tasks.
Key Entities
DeepSeekDeepSeek V4 FlashDeepSeek V4 ProOpenAI GPT-5.2Gemini 3.0 ProMixture-of-ExpertsLarge Language Model