TechFlow reports that on September 10, the B.AI platform announced the official launch of DeepSeek's newly released native multimodal MoE large model, DeepSeek-V4.1-Flash. The model features a 552 billion parameter backbone and a pioneering Causal Encoder-Decoder architecture, utilizing asymmetric computational scales for input and output to significantly reduce operational costs while comprehensively surpassing the previous generation flagship model. Designed with support for a 1 million token context window and an ultra-large 384K output length, it is specifically tailored for long-horizon programming agents, multimodal workflows, and high-concurrency agent systems. Access via both Web Chat and API is now open, allowing developers to effortlessly experience this highly cost-effective multimodal large model. More major perks are coming soon; please stay tuned to the official B.AI updates.
Add to Favorites
Share to Social Media





