News
Musk said that the 2 trillion parameter large model is about to complete training, or it may challenge the performance of Kimi K3. After Dark Side of the Moon released a new generation of open source large model Kimi K3, Tesla CEO Musk said that its 2 trillion parameter large model is expected to complete preliminary training next week, and said that the overall performance of the model will exceed the previous 1
2 min read
Source: Telegram AI频道
Musk said that the 2 trillion parameter large model is about to complete training, or it may challenge the performance of Kimi K3. After Dark Side of the Moon released a new generation of open source large model Kimi K3, Tesla CEO Musk said that its 2 trillion parameter large model is expected to complete preliminary training next week, and said that the overall performance of the model will exceed the previous 1.5 trillion parameter version, and may even surpass Kimi K3. Previously, Musk commented that Kimi K3 was "impressive" in the comment area of relevant evaluation reports. On July 18, he further revealed the progress of new model training, which triggered the industry's attention to the competition of ultra-large-scale models. It is reported that Kimi K3 has a parameter scale of 2.8 trillion, a context window of 1 million words, and supports native visual understanding capabilities. It is currently one of the open source large models with the largest parameter scale in the world. On the day the model was released, it ranked first in the world in the Arena AI front-end programming rankings with 1679 points, surpassing mainstream overseas models such as Claude Fable5 and GPT-5.6Sol. As the competition for large models enters a new stage, parameter scale, reasoning capabilities, context length, and open source ecology have become important directions for manufacturers to compete. Musk's AI company xAI has continued to promote the research and development of large models in recent years. This announcement of the training progress of the 2 trillion parameter model also shows that global AI companies are engaging in more intense technical competition around the next generation of basic models. via AI News (author: AI Base)