New AI model cuts inference costs by over 50%
DeepSeek AI has rolled out DeepSeek-V3.2-Exp, a new model that cuts inference costs by more than half compared to its previous version.
Thanks to its clever DeepSeek Sparse Attention (DSA) system, the model focuses only on the most important parts of your input—making it two to three times quicker with long documents and using up to 40% less memory.
DSA is a game-changer for processing large files
DSA is a big deal for anyone dealing with huge files—think analyzing long documents or sifting through massive codebases.
Old-school models slow down as files get longer, but DSA's "lightning indexer" keeps things speedy by zeroing in on what matters most.
The open-source nature of the model encourages innovation
With lower costs and faster responses, AI tools powered by DeepSeek-V3.2-Exp could soon be more accessible for startups, students, and developers alike.
Plus, since it's open-source under the MIT license, anyone can build on it—opening doors for fresh ideas across tech and beyond.