DeepSeek officially announces the upgrade of the online model, version number DeepSeek-V3.1-Terminus.

DeepSeek announced this evening that their online model has been upgraded. The current version is DeepSeek-V3.1-Terminus, which includes two versions: a thinking model and a non-thinking model, both with a context length of 128k. Users can now experience it online. The default output length for the non-thinking model is 4K, with a maximum of 8K, while the default output length for the thinking model is 32K, with a maximum of 64K. In terms of pricing, the model charges 0.5 RMB per million tokens for input, 4 RMB for cache misses, and 12 RMB for output.