U.S. AI Labs Release Competing Open-Source Models to Challenge China’s Dominance
- Deep Cogito released Cogito v2.1, a model with 671 billion parameters, claiming it as the best open-weight LLM by a U.S. company.
- The Allen Institute for AI introduced Olmo 3, emphasizing full transparency in training data and code.
- Cogito v2.1 is based on the Chinese DeepSeek model but offers efficiency improvements with shorter reasoning chains.
- Olmo models are built from scratch with a new dataset and offer three variants with up to 32 billion parameters.
- Deep Cogito secured $13 million in seed funding led by Benchmark, while Nvidia supported Olmo’s development.
The release of these models highlights the U.S.’s efforts to compete in the open-source AI space dominated by Chinese models like DeepSeek and Qwen. The contrasting approaches—Cogito’s rapid iteration on existing Chinese models versus Olmo’s complete transparency—illustrate different strategies for achieving technological independence.
These developments underscore the importance of maintaining control over AI infrastructure, as reliance on foreign technology could compromise U.S. firms’ autonomy and bargaining power in shaping industry standards.(Source)
