deepseek-ai/DeepSeek-V4-Flash-0731
DeepSeek’s new V4 Flash model is being pitched as unusually strong for its price.
Simon Willison notes it is a 304B-parameter release with “substantially enhanced agentic capabilities.” Artificial Analysis ranks it ahead of MiniMax M3, despite that model being larger. At $0.14 per million input tokens and $0.27 per million output tokens, Willison says it may be the best value-per-intelligence model available. His own image test improved when he raised the reasoning level to high. Simon Willison's note
Simon Willison notes it is a 304B-parameter release with “substantially enhanced agentic capabilities.” Artificial Analysis ranks it ahead of MiniMax M3, despite that model being larger. At $0.14 per million input tokens and $0.27 per million output tokens, Willison says it may be the best value-per-intelligence model available. His own image test improved when he raised the reasoning level to high. Simon Willison's note
score 8