DeepSeek is testing a new version of its language model, accessible through its website and official app. The main feature of this model is its large context window, extended to 1 million tokens.

Testing Details

The testing phase is currently underway with a limited group of users. The model includes an updated knowledge base, suggesting improvements in the ability to provide relevant and accurate responses. Some users speculate that it is a Lite version of the V4 model.

For those evaluating on-premise deployments, there are trade-offs to consider carefully. AI-RADAR offers analytical frameworks on /llm-onpremise to support these evaluations.