semiconductor / news / / Business Today
DeepSeek's V4 family is the company's most ambitious release since its R1 model rattled global markets in early 2025.
DeepSeek's V4 models are optimised for Huawei chips, not Nvidia GPUs, for the first time.
KEY POINTS
- V4-Pro activates only 49 billion parameters per token, cutting compute costs substantially.
- V4-Pro pricing is up to 50 times cheaper than comparable models like Claude Opus.
- Huawei announced full support for DeepSeek V4 across its Ascend chips after the launch.
- V4-Pro offers a 1 million token context window, a significant increase over previous models.
COMPANIES
Summarized by Newsio from Business Today. How we summarize →