机器之心 (微信公众号)T2·68 pts
- Proposes 'Collective Model Intelligence,' positioning models themselves as the fourth growth factor alongside compute, algorithms, and data.
- Mio 1.0 Lite achieved the highest average score (77.15/100) across nine public benchmarks while maintaining the lowest cost per request (0.06 CNY).
- The framework outperformed runners-up on five datasets and reduced costs by 73.7%–97.5% compared to alternative solutions.