Procurement

Post Product

  • Post Supply
  • Manage Supplies

Moore Threads Validates 2.8-Trillion-Parameter Kimi Training on Domestic GPU

   2026-11-18 Forbes / NetEase70

Moore Threads says its lab has completed a 400-step training validation of Moonshot AI's Kimi K3.1 Pro model on a domestic GPU cluster, an advance for China's AI chip ecosystem.

On November 17, Moore Threads announced that its laboratory had completed adaptation of the training pipeline for Moonshot AI's Kimi K3.1 Pro model, running 400 consecutive steps on a two-server, 16-card configuration without numerical overflow or abnormal interruption.

Kimi K3.1 Pro is a 2.8-trillion-parameter mixture-of-experts native multimodal model whose dynamic sparse attention and expert routing place high demands on multi-card communication and memory management. The test used a trimmed configuration - 13 language layers, 16 routing experts with top-eight activation, and a complete 27-layer vision tower - trained from random initialisation on a visual question-answering dataset. Training loss fell from 8.9365 to 2.1802, with peak memory use of 44.80 GiB.

The company stressed the work is a pipeline verification under a trimmed configuration, not a full-scale pre-training run. The announcement comes as domestic GPUs accelerate adaptation to leading Chinese models. TrendForce forecasts China will ship about five million high-end AI chips in 2026, up 83 percent, with domestic solutions nearing a 90 percent share.

 
ReportCollect0Reward 0Comment 0
Disclaimer
• 
This article is an original work by {author}. Reproduction is welcome, but please indicate the original source: {linkurl}. The views expressed in this article are those of the author alone, and the website has not verified the content. Readers are advised to use it for reference only. If the article involves content that violates public morality or laws, it will be deleted immediately upon discovery, and the author shall bear the corresponding responsibilities. In case of copyright or other issues, please contact us in a timely manner.
 
More>Similar News
Recommended images and text
Recommend News
Click to rank

Sign Up

For enterprise

Tel:132-7094-5888

Scan with Phone

Support

Tel:132-7094-5888

Program

Mini Program

Procurement Advisory

Scan with WeChat

WeChat

Business Opportunities

Scan with WeChat

Top