DeepSeek releases V4 Pro and Flash models with 1M context, runs on Huawei chips

DeepSeek has released its new V4 family of models, including V4 Pro and V4 Flash, which boast a 1 million token context window. These models were trained on 32 trillion tokens and feature a novel hybrid attention system for improved efficiency. Notably, the V4 Pro is designed for complex tasks, while V4 Flash offers a faster alternative, and both are released under an MIT license, with compatibility for Huawei Ascend chips. AI

影响 Advances open-weight long-context and agentic coding performance, potentially challenging closed frontier models and enabling more complex AI applications.

排序理由 Release of a new major model family from a significant AI lab with advanced capabilities like 1M context window and novel attention mechanisms.

在 Latent Space (swyx) 阅读 →

• Claude 4.6
• CUDA
• DeepSeek
• DeepSeek V4 Flash
• DeepSeek V4 Pro
• GPT-5.4
• Huawei Ascend
• Kimi K2.6
• MIT license
• NVIDIA
• Xiaomi Mimo 2.5

• 模型发布
• 论文
• 基础设施

AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →