# DeepSeek V4: Chinese Open-Source Model Rivals Frontier AI at 1/6th Cost

> DeepSeek is back with V4, and the headline is not just quality but licensing. The release comes in two open-weight sizes, V4-Pro at 1.6 trillion parameters with 49 billion active and V4-Flash at 284 billion with 13 billion active, both under an MIT license that permits commercial use and fine-tuning, with weights on Hugging Face and a 1 million token context window. DeepSeek says it rivals closed frontier models on quality, and that quantized versions run locally on a 128GB Apple Silicon machine. The engineering is the interesting part. A new hybrid attention mechanism reportedly cuts FLOPs by 27% and the KV cache by 90% compared with DeepSeek-V3.2, which is what makes running something this large on a laptop even plausible. The strategic subtext is louder still. An openly licensed model claiming frontier quality at a fraction of the cost reignites the debate about whether China has closed the gap with US labs, and it puts real pressure on anyone charging premium prices for closed weights. Older DeepSeek models retire fully on July 24.

_Section: [Daily AI Updates](https://www.wortins.com/daily-ai) · Source: Lambda · Published Tuesday, July 14, 2026_

## Wortins' read

DeepSeek is back with V4, and the headline is not just quality but licensing. The release comes in two open-weight sizes, V4-Pro at 1.6 trillion parameters with 49 billion active and V4-Flash at 284 billion with 13 billion active, both under an MIT license that permits commercial use and fine-tuning, with weights on Hugging Face and a 1 million token context window. DeepSeek says it rivals closed frontier models on quality, and that quantized versions run locally on a 128GB Apple Silicon machine. The engineering is the interesting part. A new hybrid attention mechanism reportedly cuts FLOPs by 27% and the KV cache by 90% compared with DeepSeek-V3.2, which is what makes running something this large on a laptop even plausible. The strategic subtext is louder still. An openly licensed model claiming frontier quality at a fraction of the cost reignites the debate about whether China has closed the gap with US labs, and it puts real pressure on anyone charging premium prices for closed weights. Older DeepSeek models retire fully on July 24.

## Source

[Read the full story at Lambda](https://lambda.ai/blog/deepseek-v4-the-most-expected-open-source-model)

## Related coverage

- [Google launches Africa's first Applied AI Lab in Accra for African founders and researchers](https://www.wortins.com/story/google-launches-africa-s-first-applied-ai-lab-in-accra-for-a-2f00b44d) — [Further Africa](https://furtherafrica.com/2026/07/13/google-africa-ai-lab-launches-in-accra-ghana/)
- [xAI open-sources Grok Build coding agent](https://www.wortins.com/story/xai-open-sources-grok-build-coding-agent-7997c3dd) — [Future Tools](https://futuretools.io/news)
- [Andrej Karpathy and Tom Blomfield Join Anthropic's Team](https://www.wortins.com/story/andrej-karpathy-and-tom-blomfield-join-anthropic-s-team-0d91a963) — [unrot.co](https://unrot.co/blogs/today-top-10-ai-news-july-15-2026/)
- [Lovable Raises $200M Series B at $2.8B Valuation for No-Code App Generation](https://www.wortins.com/story/lovable-raises-200m-series-b-at-2-8b-valuation-for-no-code-a-fd17a524) — [blog.mean.ceo](https://blog.mean.ceo/ai-startup-funding-news-july-2026/)
- [Apple Intelligence approved for China, uses Alibaba Qwen models](https://www.wortins.com/story/apple-intelligence-approved-for-china-uses-alibaba-qwen-mode-808e0ae5) — [Build Fast with AI](https://www.buildfastwithai.com/blogs/ai-news-today-july-16-2026)
- [South Korea announces 1,350 trillion won ($880B) decade-long AI infrastructure plan](https://www.wortins.com/story/south-korea-announces-1-350-trillion-won-880b-decade-long-ai-dd479678) — [Tech Startups](https://techstartups.com/2026/07/16/top-tech-news-today-july-15-2026/)

---

_Curated and written by [Wortins](https://www.wortins.com) — The daily AI briefing. Every story links to its original source; the "Wortins read" on each is our own original analysis. [About Wortins & our editorial approach](https://www.wortins.com/about)._