Abstract:
On September 12, according to feedback from multiple users, DeepSeek is currently conducting a grayscale test of voice dialogue, and a small speaker button will appear in the upper right corner of the DeepSeek App for some users.
At the same time, the "Reading Tone" option will appear in the voice bar of DeepSeek's settings page. Users can choose from the following four timbres: Shell (various and lively), Bailang (bright and firm), Starfish (playful and sweet), Undercurrent (low and deep).
According to DeepSeek news, the DeepSeek V4.1 Flash model was officially released on September 10. This is the smallest model in DeepSeek’s new model structure series, with native multi-modal visual understanding capabilities. The original design intention of the new model structure is to have a higher upper limit of capabilities, faster inference speed, greater throughput, and scalability to larger parameter models.
DeepSeek stated that the DeepSeek V4.1 Flash model successfully surpassed the intelligence level of a number of flagship models including DeepSeek V4 Pro in benchmark tests.
Comments