Large Models Return to Parameter Race: 2 Trillion Becomes Standard, 3 Trillion Next? - Sina Finance
Large AI models are returning to a focus on parameter count as a key competitive metric, with 200 billion parameters becoming a standard for top models and speculation that 300 billion could be the next milestone. Major Chinese AI companies are reportedly scaling up their training parameters to stay competitive.
The article from Sina Finance discusses a renewed competition among large language models centered on the number of parameters. It notes that models with around 200 billion parameters have become the industry standard for leading players, and there is speculation that the next step could involve models reaching 300 billion parameters. The piece suggests that this trend indicates a return to scaling as a primary strategy for improving model capability, with implications for computational resources and training costs.
Source: Moonshot Kimi (GNews) —
original