After the release of its new open-source large model, Kimi K3, Moonshot AI, the company behind it, saw Elon Musk, CEO of Tesla, state that its 2-trillion parameter-scale model is expected to complete initial training by next week, and he claimed that the overall performance of this model would surpass its previous 1.5-trillion parameter version, and may even exceed Kimi K3.

Prior to this, Musk had commented on the evaluation reports related to Kimi K3, calling it "impressive." On July 18, he further revealed the progress of the new model's training, sparking industry attention on the competition for ultra-large-scale models.

It is known that Kimi K3 has a parameter scale of 2.8 trillion, a context window of up to 1 million words, and supports native visual understanding capabilities, making it one of the largest open-source large models in the world. On the day of its release, it topped the global ranking on the Arena AI front-end programming leaderboard with 1679 points, surpassing overseas mainstream models such as Claude Fable5 and GPT-5.6Sol.

As the competition for large models enters a new phase, parameter scale, reasoning capability, context length, and open-source ecosystems have become key areas of contention among manufacturers. xAI, the AI company under Musk's leadership, has been continuously advancing large model R&D in recent years. The recent announcement of progress on its 2-trillion parameter model also highlights the increasingly intense technological competition among global AI companies for the next-generation foundational models.