AI BriefingDeepSeek Unveils V4 Model with Ultra-Low-Price Strategy
Key point
DeepSeek has unveiled the V4 model, leveraging Huawei chips to boost performance while dramatically lowering prices.
Details
Chinese AI startup DeepSeek has unveiled its latest large language model, V4. The model was released in two versions: DeepSeek-V4-Pro (1.6 trillion parameters, 1 million token context) and the smaller, cheaper DeepSeek-V4-Flash.
The V4 models claim to overwhelm existing open source models in agentic coding and reasoning. Benchmark results show performance similar to GPT-5.4 and Gemini 3.1 Pro, narrowing the gap with state-of-the-art frontier models to roughly 3-6 months.
The most striking aspect is its dramatic pricing policy. V4-Pro's output cost per 1 million tokens is $3.48, overwhelmingly cheaper compared to OpenAI ($30) or Anthropic ($25). V4-Flash costs just $0.28, signaling a price disruption in the market.
This model was trained using Huawei's Ascend AI processors, suggesting China's AI development capabilities and the possibility of hardware self-sufficiency. Following this news, the market reacted strongly, with shares of Chinese semiconductor manufacturer SMIC surging.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Toss Doctor V2 Released: Maximizing QA Efficiency with LLM-based Automated Testing and Self-healing