Micron Tech Return Rate
Likes ReceivedBreaking: DeepSeek has released a new version after a year, with its most powerful open-source large model, DeepSeek V4, officially launching.
This new model comes in two versions: V4 Pro and V4 Flash.
V4 Pro excels in evaluations of reasoning ability, world knowledge, mathematics, STEM fields, and competition-level code. The official statement indicates its performance has reached a level comparable to top-tier closed-source AI models in the US.
V4 Flash focuses on speed and cost advantages, making it suitable for small companies or individual developers with limited budgets. It can achieve results close to V4 Pro in simple Agent tasks.
DeepSeek V4 employs a new attention mechanism and DSA sparse attention technology, significantly reducing the computational load and VRAM requirements for long contexts.
From today onwards, the official service will offer 1M (one million characters) context as standard. In terms of API pricing, V4's price is only 1/10 to 1/50 of GPT 5.4 and Claude Opus 4.6, while being fully open-source.
Overall, this is a release of an open-source model that is affordable, powerful in performance, and excellent in context handling capabilities.
The copyright of this article belongs to the original author/organization.
The views expressed herein are solely those of the author and do not reflect the stance of the platform. The content is intended for investment reference purposes only and shall not be considered as investment advice. Please contact us if you have any questions or suggestions regarding the content services provided by the platform.
