Next Article
DeepSeek's new AI model can cut your API costs
Technology
DeepSeek just dropped its latest AI model, V3.2-exp, and it's all about making things cheaper for anyone working with huge chunks of data.
Thanks to its "Sparse Attention" tech—a more efficient way to process long-context data—the model can process loads of data without putting too much strain on servers.
Model is available on Hugging Face
Early tests suggest V3.2-exp could cut API costs by up to 50%, which is a pretty big deal if you're building apps or services that rely on AI.
DeepSeek, a Chinese company known for experimenting with new training methods, is keeping things open-weight and freely available on Hugging Face—so expect lots of third-party testing soon.