DeepSeek Releases V4.1 Flash 1017 Model and Opens Weights
DeepSeek-V4.1-Flash-1017 is now available. The official release notes highlight improvements in reasoning, coding, agent tasks, and visual evaluation. It supports a 1 million token context window, is released under the MIT License, and includes instructions for local inference.
DeepSeek-V4.1-Flash-1017 has been released. This is a minor version update to DeepSeek-V4.1-Flash. According to the official announcement, it has seen improvements in reasoning, programming, agent tasks, and visual evaluation, while also reducing task execution time and output length, thereby lowering usage costs.
The model supports a 1 million token context window, is licensed under the MIT License, and includes instructions for local inference.
DeepSeek | Huggingface
https://mp.weixin.qq.com/s/X-5kbBkedQ5Oct9byUunag
https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017
