Key Highlights
- Kimi K3 model weights become freely accessible for public download starting Monday, July 27
- With 2.8 trillion parameters, K3 stands as the largest open-weight artificial intelligence model globally
- Moonshot experienced a sixfold increase in daily revenue following K3’s initial launch this month
- White House officials have alleged the company used prohibited Nvidia hardware for K3 training
- The startup pursues new financing at a $50 billion valuation while preparing for a Hong Kong stock exchange listing
The Beijing-headquartered artificial intelligence company Moonshot is opening access to its Kimi K3 model, enabling developers worldwide to download, customize, and deploy the technology without charge.
Starting Monday, the Chinese AI firm will distribute the complete model weights publicly. This move empowers software developers to obtain, adapt, and independently operate the system on their own infrastructure.
Boasting 2.8 trillion parameters, Kimi K3 claims the title of the planet’s most expansive open-weight artificial intelligence system. The parameter count serves as an indicator of the model’s sophistication and capabilities.
K3 incorporates an expansive 1-million-token context capacity, enabling it to analyze extensive documents or comprehensive software repositories within one interaction.
According to Moonshot’s assessments, K3 demonstrates performance on par with OpenAI’s Fable 5 while surpassing Anthropic’s Opus offering and multiple GPT-5 iterations across various evaluation metrics.
China’s AI Sector Advances
The introduction of K3 has sparked parallels to DeepSeek’s debut in early 2025. Both releases demonstrated that Chinese technology firms can develop competitive artificial intelligence systems using reduced resources and smaller budgets compared to American counterparts.
AI models originating from China have expanded their worldwide user base by providing more economical pricing while maintaining similar capabilities. Leading cloud service platforms such as Amazon Web Services, Microsoft Azure, and Google Vertex AI have yet to integrate these Chinese open-weight alternatives into their offerings.
Analysts at Bloomberg Intelligence observed that while adoption of open-weight models continues to climb, revenue generation from these systems hinges on securing distribution agreements with major cloud infrastructure providers.
Investment Rounds, IPO Strategy, and Regulatory Challenges
Following K3’s public debut, Moonshot witnessed daily sales multiply by at least six times. The company briefly suspended new customer registrations to manage the overwhelming demand.
The startup achieved a $20 billion company valuation this May after securing $2 billion in investment led by Long-Z Investments, the venture capital division of Meituan. Currently, Moonshot is negotiating additional funding that would establish a $50 billion valuation for the three-year-old enterprise.
Plans for a Hong Kong stock market debut could materialize within the current year. Company founder Yang Zhilin has publicly stated that Moonshot’s growth strategy emphasizes transparency and open access rather than closed, proprietary frameworks.
The organization has drawn scrutiny from United States authorities. Last week, a White House representative claimed Moonshot developed K3 using restricted Nvidia processors and employed distillation techniques ā a method involving extraction of outputs from competing models to enhance proprietary systems.
Moonshot has yet to address these claims publicly.
An in-depth technical document detailing K3’s design framework, training procedures, and performance evaluations is anticipated to accompany the model’s public release.


