← Latest news 
DeepSeek V4 slashes AI costs to one sixth of GPT 5.5 yet nears frontier intelligence
Technology
Published on 24 April 2026
Open MIT weights plus a 1 million token window
DeepSeek-V4 has arrived as a free, MIT-licensed 1.6T Mixture-of-Experts model that reportedly matches or beats top closed systems on select benchmarks while costing about one-sixth as much as GPT-5.5 via API. The bigger story: a native one-million-token context achieved with new attention and training techniques, pressuring premium model pricing.
- DeepSeek-V4-Pro is priced around one-sixth to one-seventh GPT-5.5 on typical API usage
- The model offers a native 1M-token context window with sharply reduced KV cache needs
- Benchmarks show mixed results, but DeepSeek is close on several practical agent and coding tests
- Weights are MIT-licensed, supporting commercial use and local deployment
Read the full story at Venture Beat
This summarization was done by Beige for a story published on 
Venture Beat