TechCrunch: Study Suggests DeepSeek and Moonshot Model Weights Exhibit Blackwell Training Characteristics
According to TechCrunch, Will DePue of the non-profit research organization Unemployment Labs analyzed the public model weights of DeepSeek, Moonshot AI, MiniMax, and Ant Group. He reported finding signatures indicating training on NVIDIA Blackwell chips in models such as Moonshot's Kimi K3 and DeepSeek V4. The report also noted that this does not imply the models were trained exclusively on NVIDIA chips, nor does it necessarily indicate a violation of export controls, as the relevant compute infrastructure may be located in overseas data centers.
No AI analysis yet. Tap the "AI Analysis" button above to generate one now.
Disclaimer: This content reflects the author's personal views only and does not constitute investment advice.
24H Trending
-
1
Citadel Securities has poached Matthew Zimmerman from Diameter Capital Partners to head investment-grade corporate bond trading.
-
2
What is RCN coin? Where to trade and buy RCN coin?
-
3
What is HNW coin? Where can HNW coin be traded and bought?
-
4
Microsoft services suffer widespread outage impacting thousands of users in North America; Amazon Web Services and Google users also report disruptions
-
5
September Ethereum Price Outlook: Can Polymarket's $2,600 Target Be Achieved?
-
6
How Will BlackRock's Staked Ethereum ETF Impact the Ethereum Ecosystem on Its First Day with $15.5 Million in Inflows?
-
7
What is BFT token? Can BFT token be held for the long term?
-
8
CNBC: NVIDIA Drops 11% This Week, While AMD, Broadcom, and TSMC Rise
-
9
What is the value of OLIVE coin? Is OLIVE coin a good long-term investment?
-
10
Chinese stocks led a rally in Asian chip shares as investors shrugged off the NVIDIA controversy surrounding Jensen Huang and rushed back into AI trades.
Markets Today
Recommended Reading





