US EditionChina's DeepSeek says its hit AI model cost just $294,000 to train
DeepSeek trained its R1 AI model using 512 Nvidia H800 chips, achieving a training cost of $294,000, significantly lower than U.S. competitors' expenses.
- DeepSeek, a Chinese AI developer based in Hangzhou, released its lower-cost R1 AI model in January, which cost $294,000 to train.
- This release came after U.S. export regulations, enacted in October 2022, prohibited Nvidia from exporting its high-end H100 and A100 AI processors to China, while DeepSeek reportedly obtained Nvidia-designed H800 chips intended for that market to support its training efforts.
- The R1 model trained for 80 hours on a 512-chip cluster of Nvidia H800 GPUs, and researchers used A100 GPUs only in initial preparatory phases.
- A peer-reviewed article in Nature detailed these training costs and setups, noting the $294,000 figure is much lower than the over $100 million U.S. rival OpenAI's CEO Sam Altman reported for foundational model training.
- DeepSeek's claims and technology use have been questioned by U.S. officials and companies, and the release prompted investor concerns about competition with established AI leaders.
8 Articles
8 Articles
China's DeepSeek says its hit AI model cost just $294,000 to train
Chinese AI developer DeepSeek said it spent $294,000 on training its R1 model, much lower than figures reported for U.S. rivals, in a paper that is likely to reignite debate over Beijing's place in the race to develop artificial intelligence.
China's Deepseek Says Its Hit AI Model Cost Just $294,000 To Train
Chinese AI developer DeepSeek said it spent $294,000 on training its R1 model, much lower than figures reported for US rivals, in a paper that is likely to reignite debate over Beijing's place in the race to develop artificial intelligence.
The Mighty 790 KFGO
Investing.comChina’s DeepSeek says its hit AI model cost just $294,000 to train
By Eduardo Baptista BEIJING (Reuters) -Chinese AI developer DeepSeek said it spent $294,000 on training its R1 model, much lower than figures reported for U.S. rivals, in a paper that is likely to reignite debate over Beijing's place in the race to dev...
Chinese artificial intelligence developer DeepSeek said he spent $294,000 on training his R1 model, a much lower figure than his US rivals, in an article that will probably rekindle the debate about Beijing’s place in the AI development race. What do costs refer to? Hangzhou’s rare update – the first estimate that makes R1’s training costs public – appeared in a peer-reviewed article in the academic journal Nature .
Coverage Details
Bias Distribution
- 57% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
