DeepSeek V4 Now Generally Available: 1M Context Window, MIT-Licensed Open Weights, and Industry-First Surge Pricing

What's new

• DeepSeek V4-Pro: 1.6T total parameters / 49B active (MoE), 1M token context window
• DeepSeek V4-Flash: 284B total / 13B active -- faster and cheaper variant for production deployment
• MIT license: open-weight release -- commercially usable without restrictions, including fine-tuning
• Surge pricing (first in AI): 2x higher API rates during peak hours (9:00--12:00 and 14:00--18:00 Beijing time)
• Alias retirement: deepseek-chat and deepseek-reasoner aliases retired on July 24

Why it matters

DeepSeek V4-Flash (~$0.02/task) is dramatically cheaper than Claude/GPT at comparable quality -- a direct commercial threat. MIT licensing enables unrestricted local deployment. Surge pricing sets a precedent that may ripple through the AI API industry.

How to try it

API access via platform.deepseek.com -- migrate from aliases to explicit versions (deepseek-v4-pro, deepseek-v4-flash). Open weights available on Hugging Face.