At the Amazon Web Services China Summit, Moonshot AI announced the latest advancements of Kimi. Data shows that Kimi’s overseas paid users and API revenue have grown by 400% year-over-year, with services now available in over 200 countries and regions across industries including internet, finance, manufacturing, education, and healthcare. The company maintains a lean team of 300 employees focused exclusively on foundational model development. On the technical front, the latest K2.7Code model has improved performance by 10% to 30% over its predecessor, bringing its programming and logical reasoning capabilities closer to industry-leading standards. Kimi plans to launch the K3 model in Q3, with an expected parameter scale of 2.52 trillion and an extended context window of up to 1M, alongside comprehensive enhancements to multimodal processing capabilities—demonstrating its ambition to compete with the world’s top AI leaders.Article author and source: AWS China Summit
At the recent Amazon Web Services China Summit, Chinese large model startup Moonshot AI revealed the latest progress of its core product, Kimi. Data shows that Kimi’s overseas paying users and API revenue have both experienced 400% year-over-year growth, with the product now reaching over 200 countries and regions worldwide and serving key industries including internet, finance, manufacturing, education, and healthcare.
Amid today’s intense competitive landscape, Moonshot AI has firmly prioritized R&D as its core strategic focus. The company maintains a lean team of approximately 300 employees and channels its resources primarily toward foundational model development, rather than mere delivery tasks. Regarding its vision, Moonshot AI’s B2B lead, Huang Zhenxin, stated that Kimi aims to directly compete with the world’s leading forces in AI and secure technological influence.
In terms of technological iteration, Kimi's evolution rate is equally noteworthy. Its latest K2.7Code model has improved performance by 10% to 30% over the previous generation, bringing its programming and logical processing capabilities even closer to industry-leading standards.
Even more exciting, the company plans to launch its next-generation large model, K3, in the third quarter of this year. According to reports, the model’s parameter scale is expected to surge to 2.52 trillion, the context window will be expanded to 1M, and its multimodal processing capabilities will be significantly enhanced. Amid the rapid emergence of domestic large models, Kimi’s intensive iterations and scaling demonstrate its ambition and confidence to compete with the world’s top-tier models.