Chinese AI start-up Moonshot AI has launched Kimi K3, a 2.8 trillion-parameter model built for coding, knowledge work and complex reasoning. The company says the open-weight system can compete with leading models from OpenAI and Anthropic, while its full model weights are scheduled for release on July 27.
Parameters are internal variables learned during training and can provide a rough indication of a model’s scale, though a higher count does not automatically produce better performance. Moonshot describes Kimi K3 as the first open-weight model to approach the three-trillion-parameter level.
Built for Long-Running Tasks
According to Moonshot’s official Kimi K3 announcement, the model has a one-million-token context window and supports both text and images. It is designed to handle long-running software development and engineering tasks with limited human supervision.
Unlike proprietary models whose underlying weights remain private, Kimi K3’s open-weight release will allow developers to download, run and customise the system. Its size, however, means that self-hosting could require computing equipment costing hundreds of thousands of dollars.
Moonshot is backed by Chinese technology groups Alibaba and Tencent. The company described Kimi K3 as its “most capable flagship model to date,” as it expands its position in China’s generative AI market.
Kimi K3 Leads Coding Evaluation
Independent evaluations indicate that Kimi K3 performs close to several leading US models. Arena ranked it first in an assessment of web interface development, where human evaluators selected between outputs without knowing which model produced them.
Kimi K3 also placed second overall in an evaluation by Vals AI, behind Anthropic’s Fable 5 and ahead of OpenAI’s GPT-5.6 Sol. Artificial Analysis found its performance comparable to OpenAI’s GPT-5.5 and Anthropic’s Claude Opus 4.8 on tests involving complex, multi-step tasks.
Moonshot said Kimi K3 performed competitively with Fable 5 on GPU optimisation and surpassed Claude Opus 4.8, GPT-5.6 Sol and GPT-5.5 in selected tests. The company’s results have not yet been fully verified using the downloadable model weights.
Chinese AI Stocks Fall After Launch
The announcement affected shares of competing Chinese AI companies listed in Hong Kong. Zhipu, also known as Z.ai, fell about 27.7% shortly before the market closed, while MiniMax declined roughly 16.5%, according to Reuters.
The launch follows US restrictions on the availability and export of some advanced AI systems and hardware. Moonshot developed Kimi K3 despite US controls limiting China’s access to certain high-end semiconductors.
Featured image credits: Codersera
For more stories like it, click the +Follow button at the top of this page to follow us.