deepseek-ai/DeepSeek-V4-Flash-0731
DeepSeek has released DeepSeek-V4-Flash-0731, the latest in its V4 family, described as having "substantially enhanced agentic capabilities." The model has 304 billion parameters and is 167GB on Hugging Face. Despite its size, it appears to perform strongly, with Artificial Analysis ranking it ahead of MiniMax M3, a 428B model. Its pricing of $0.14 per million input tokens and $0.27 per million output tokens may make it the best value-per-intelligence model currently available, according to the source. It also performs well on the Intelligence Index vs. Cost per Intelligence Index Task chart. Simon Willison tested the model via OpenRouter and initially got a disappointing pelican image using the default reasoning level, but got a much better result when he bumped the reasoning level to high using the command: llm -m openrouter/deepseek/deepseek-v4-flash-0731 -t pelican -o reasoning_effort high.
DeepSeek-V4-Flash-0731 offers strong performance at a low price, potentially making it a cost-effective choice for developers.