DeepSeek V4 - almost on the frontier, a fraction of the price

📰 Simon Willison's Blog

DeepSeek releases V4 series with two preview models, offering a fraction of the price of similar models, learn how to utilize them

intermediate Published 24 Apr 2026

Action Steps

Explore the DeepSeek-V4-Pro and DeepSeek-V4-Flash models
Compare their performance with other similar models
Configure the models for specific tasks using the 1 million token context Mixture of Experts
Test the models' efficiency and accuracy in various applications
Apply the models to real-world problems to evaluate their potential

Who Needs to Know This

AI engineers and researchers can benefit from this new release to improve their models' performance and efficiency, while product managers can explore the potential applications of these models

Key Insight

💡 DeepSeek's V4 series offers a cost-effective solution for AI models with high performance and efficiency