China's Moonshot AI has released Kimi K3, an open-source model with 2.8 trillion parameters — the largest such model ever made public. The launch comes as Anthropic's AI model continues to lead user ratings, holding a 91.5% 'YES' score as of August 2026.
A New Scale in Open-Source AI
Kimi K3's 2.8 trillion parameters dwarf previous open-source releases. Most open-source models top out at a few hundred billion parameters, making this a significant leap. Moonshot AI, a Beijing-based startup, has not disclosed the training data or compute resources used, but the sheer size suggests a massive investment in infrastructure.
The model is available for download and use under an open-source license, allowing researchers and developers to inspect, modify, and deploy it. That could accelerate progress in fields like natural language processing and code generation, where larger models often perform better on complex tasks.
Anthropic's Rating Lead
While Moonshot AI pushes the boundaries of scale, Anthropic's AI model holds a 91.5% 'YES' rating — the highest among comparable systems as of August 2026. The rating, based on user feedback or benchmark evaluations, reflects strong satisfaction with the model's outputs. Anthropic has not released a model of similar size, focusing instead on safety and alignment.
The contrast highlights two different strategies in AI development: one betting on raw scale, the other on refined performance and user trust.
What the Launch Means for Developers
Kimi K3 is now available for download from Moonshot AI's website and through major model repositories. Developers can run it on their own hardware or access it via cloud services. The open-source nature means the community can build on top of it, fine-tune it for specific tasks, or even attempt to shrink it for smaller deployments.
Early adopters will likely test Kimi K3 against established models, including Anthropic's, to see how it handles reasoning, creativity, and factual accuracy. The model's size may give it an edge in some areas, but it also requires substantial computational resources to run.
The release of Kimi K3 gives developers access to a model of unprecedented scale. How it stacks up against proprietary models like Anthropic's in real-world tasks will likely be the next focus for the AI community.
