Jinshi data news on February 26th, Fourth Paradigm (06682.HK) launched the large model inference integrated machine solution SageOneIA today, further reducing the cost of large model inference. For example, the full version of DeepSeekV3/R1 only requires two integrated machines to use. In addition, the integrated machine solution also integrates intelligent Computing Power pooling technology. On the basis of supporting mainstream large models such as DeepSeekV3/R1, QWen2.5, LLama3.3, enterprises can flexibly switch between the full version and multiple distillation models, with GPU utilization rate increased by more than 30% and inference performance increased on average by 5-10 times.
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
The Fourth Paradigm launches the large-scale model reasoning integrated solution SageOne IA
Jinshi data news on February 26th, Fourth Paradigm (06682.HK) launched the large model inference integrated machine solution SageOneIA today, further reducing the cost of large model inference. For example, the full version of DeepSeekV3/R1 only requires two integrated machines to use. In addition, the integrated machine solution also integrates intelligent Computing Power pooling technology. On the basis of supporting mainstream large models such as DeepSeekV3/R1, QWen2.5, LLama3.3, enterprises can flexibly switch between the full version and multiple distillation models, with GPU utilization rate increased by more than 30% and inference performance increased on average by 5-10 times.