DeepSeek V4: Frontier-Class Open-Weight Model with Million-Token Context at a Fraction of the Price

Summary

DeepSeek released two Mixture-of-Experts model checkpoints on April 24, 2026: V4-Pro (1.6T total parameters, 49B active) and V4-Flash (284B total, 13B active), both with 1 million token context windows under an MIT license. The pricing is dramatically competitive — V4-Flash costs $0.14 per million input tokens and $0.28 for output, undercutting even OpenAI's budget offerings, while V4-Pro runs $1.74/$3.48 per million tokens, making it "the cheapest of the larger frontier models" while maintainin