Meta and NVIDIA Engage in AI Race

The article focuses on the escalating global competition in the field of open-weight AI models, with major companies like Meta and Nvidia releasing new models, alongside rapid progress by Chinese firms. Meta unveiled the "Mistral Gemini" model, which contains 30 billion parameters and can be run locally on consumer devices, aiming to empower local systems and perform tasks directly on devices instead of relying on remote servers. The company also plans to release a more powerful model, "Mistral Spark 1.2," to expand its developer ecosystem. Nvidia has introduced new "NeMo Train" models, with expectations that larger models containing a trillion parameters will soon emerge, reflecting increased computational capabilities and costs. On the Chinese side, "Mont Shuut AI" launched the "Chem K 3" model, part of its efforts to narrow the gap with American companies, offering advantages in customization and cost-efficiency. American companies like Meta and Nvidia are defending the open release of models, warning that strict restrictions could hinder innovation; meanwhile, the widespread adoption of open models is viewed as a geopolitical aspect of the rivalry between Washington and Beijing to secure technological and strategic influence.

Size: 120%Text Size

Notice: This Is an AI-Generated Summary

Display The Full Article