DeepSeek, a Hangzhou-based Chinese AI company founded by Liang Wenfeng, said its V4.1-Flash model uses less key-value (KV) cache than the previous generation to lower cache-related costs in AI agent workloads.

DeepSeek said its official partners WorkBuddy, including CodeBuddy, and OpenCode now support V4.1-Flash, and users can select the model through the deepseek-flash setting.

CodeBuddy is described as an AI code editor.

The company also said it plans to work with the open-source community on inference support for V4.1-Flash and explore more deployment options, including potential large-scale setups with 2,000 graphics processing units (GPUs) and a storage cluster.

DeepSeek said the update would help it serve more users at lower cost and that it would pass those savings on to users.

🔗 Source: DeepSeek

Recent DeepSeek developments