DeepSeek V4 Pro and Flash: Benchmarks, Pricing, and What It Means for the AI Arms Race in 2026

# DeepSeek V4 Pro and Flash: Benchmarks, Pricing, and What It Means for the AI Arms Race in 2026

> **Quick answer:** DeepSeek released V4-Pro and V4-Flash on April 24, 2026 — one year after the original "Sputnik moment." V4-Pro scores 80.6% on SWE-bench Verified (0.2 points behind Claude Opus 4.6) and costs $1.74 per million input tokens. V4-Flash matches Claude Sonnet 4.6 capability at just $0.14 per million tokens. Both are open weights, built to run on Huawei chips, and carry a 1 million token context window. The Sputnik sequel has arrived.

DeepSeek V4 Pro Flash release benchmarks have landed, and the numbers are hard to argue with. One year after China's DeepSeek shocked the AI world with R1, the lab is back — with two models that sit within striking distance of OpenAI and Anthropic's best work, at a fraction of the cost. Here is exactly what the data shows, why Huawei's involvement matters more than most coverage acknowledges, and how this reshapes the 2026 AI arms race.

## What DeepSeek V4 Pro and V4 Flash Actually Are

DeepSeek released preview versions of both models on Hugging Face on April 24, 2026. Both are open weights — meaning developers can download, modify, and self-host them without paying API fees. Both feature a 1 million token context window, enabling entire codebases or book-length documents to be processed in a single prompt.

**V4-Pro** is the headline model: - 1.6 trillion total parameters, 49 billion active (Mixture of Experts architecture) - The largest open weights model currently available - Scores 52 on Artificial Analysis's Intelligence Index — the #2 open weights reasoning model globally, behind only Kimi K2.6