DeepSeek V4 Pro, Flash Released in Preview
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- DeepSeek released preview versions of V4 Pro and V4 Flash on April 24, 2026, about a year after its previous model "upended Silicon Valley" (per Bloomberg)
- V4 Pro trails the performance of state-of-the-art models by about 3 to 6 months, according to DeepSeek's own positioning stated in Bloomberg's lead
- DeepSeek published the V4 model paper titled "Towards Highly Efficient Million-Token Context Intelligence" on Hugging Face, signaling a focus on long-context efficiency
- Investors rotated toward Chinese chipmakers as the release intensified AI competition, per the South China Morning Post
- Nvidia is "losing out in one key market" with DeepSeek laying bare a missed opportunity, per Barron's
- Coverage converged on a US-China AI race framing, with Nikkei, Al Jazeera, France 24, the New York Times, and CNBC all positioning the release as a geopolitical challenge to US rivals
- The Livemint and The Independent headlines spotlight DeepSeek's claims that V4 beats ChatGPT and Gemini in key benchmarks — an angle most other outlets downplayed
Why it matters: DeepSeek's V4 release, timed almost exactly a year after its first viral breakthrough, extends China's open-source AI footprint and reignites competitive pressure on US frontier labs. Barron's notes the launch lays bare a missed Nvidia opportunity in China, while investors rotated into Chinese chipmakers on the news — together signaling that the post-DeepSeek compute and hardware landscape is shifting.
