Chinese AI startup DeepSeek released preview versions of its long-awaited V4 flagship on April 24, 2026, arriving just over a year after the company shocked Silicon Valley with its original breakthrough. The lineup includes V4-Pro with 1.6 trillion parameters and V4-Flash with 284 billion parameters, both open source and featuring a new Hybrid Attention Architecture.

The standout feature is a 1 million-token context window, large enough to fit entire codebases or book-length documents in a single prompt. DeepSeek claims V4-Pro outperforms Anthropic Claude Sonnet 4.5 and approaches Opus 4.5, falling only marginally short of OpenAI GPT-5.4 and Google Gemini 3.1-Pro. The company acknowledged its trajectory trails frontier Western models by roughly three to six months.

The release ratchets up competition in the open-source AI race, pressuring US labs that keep their top models proprietary. It arrives the same week OpenAI launched GPT-5.5 and Microsoft integrated Anthropic models into its security stack, underscoring how quickly the frontier is moving in 2026.