Moonshot to Launch Kimi K3 Model with 2.5 Trillion Parameters in Q3

The competition among domestic large language models is heating up once again. According to the latest reports, Moonshot's next-generation flagship model, Kimi K3, is slated for an official launch in the third quarter of this year. As a highly anticipated newcomer, every update to the Kimi series has drawn significant industry focus. The arrival of K3 is poised to raise the competitive bar for domestic LLMs even higher.
2.5 Trillion Parameters Target the Industry's Peak
In terms of core technical specifications, Kimi K3 demonstrates a substantial generational leap. Its parameter count is expected to reach 2.5 trillion, placing it among the very top tiers of current domestic models. For comparison, the recently released DeepSeek V4 Pro has approximately 1.6 trillion parameters, while Baidu's ERNIE 5.0 features around 2.4 trillion. With its 2.5 trillion parameters, Kimi K3 aims to comprehensively surpass rivals in computational power and model capacity.
Advancing the Million-Word Context Window
Beyond its massive scale, Kimi K3 further refines its signature strength in "long-context" processing. The new model's standard context length will be extended to approximately one million words. This enhancement means users can expect a more coherent and in-depth interactive experience when working with extremely long documents, complex codebases, or large-scale datasets.
Steady Momentum: Moonshot's Business and Technological Trajectory
Reviewing Moonshot's development path reveals a rapid iteration cadence. From achieving sub-second hot updates for a trillion-parameter model last September, to releasing the open-source reasoning model Kimi K2 Thinking in November, followed by the K2.5 model topping open-source charts and reaching $100 million in Annual Recurring Revenue (ARR) early this year, the team has consistently demonstrated robust engineering execution.
Very recently, on April 21st, Kimi released and open-sourced the K2.6 version, focusing on enhanced cluster processing capabilities. Merely a week later, news of K3 emerged. Moonshot's president, Zhang Yuting, has previously emphasized the team's high efficiency in resource utilization. As K3's launch approaches, the race for trillion-parameter computational supremacy is entering a critical phase.
Related article
Anthropic Expands Claude AI Coding Tools to Japan in Push for Overseas Growth
Anthropic, a prominent U.S. artificial intelligence firm, is intensifying its global outreach. On Wednesday, the company hosted a major developer gathering, "Code with Claude," in Tokyo, drawing close to 500 software engineers. This initiative seeks
India’s Software Giant Cuts Hiring, Promises No Layoffs as AI Agents Scale Up
As artificial intelligence reshapes traditional labor-intensive business models, Tata Consultancy Services (TCS), a premier Indian software outsourcing firm, has unveiled its strategic response. During Tuesday’s annual general meeting, the chairman o
China Locks AI Models During Gaokam to Block Instant Homework Help
With the 2026 Gaokao fast approaching, rumors regarding the suspension of AI tools during the exam period have ignited intense online debate. In response to public concern, major AI platforms and educational apps have clarified their stance: rather t
Related Special Topic Recommendations
Comments (0)
0/500

The competition among domestic large language models is heating up once again. According to the latest reports, Moonshot's next-generation flagship model, Kimi K3, is slated for an official launch in the third quarter of this year. As a highly anticipated newcomer, every update to the Kimi series has drawn significant industry focus. The arrival of K3 is poised to raise the competitive bar for domestic LLMs even higher.
2.5 Trillion Parameters Target the Industry's Peak
In terms of core technical specifications, Kimi K3 demonstrates a substantial generational leap. Its parameter count is expected to reach 2.5 trillion, placing it among the very top tiers of current domestic models. For comparison, the recently released DeepSeek V4 Pro has approximately 1.6 trillion parameters, while Baidu's ERNIE 5.0 features around 2.4 trillion. With its 2.5 trillion parameters, Kimi K3 aims to comprehensively surpass rivals in computational power and model capacity.
Advancing the Million-Word Context Window
Beyond its massive scale, Kimi K3 further refines its signature strength in "long-context" processing. The new model's standard context length will be extended to approximately one million words. This enhancement means users can expect a more coherent and in-depth interactive experience when working with extremely long documents, complex codebases, or large-scale datasets.
Steady Momentum: Moonshot's Business and Technological Trajectory
Reviewing Moonshot's development path reveals a rapid iteration cadence. From achieving sub-second hot updates for a trillion-parameter model last September, to releasing the open-source reasoning model Kimi K2 Thinking in November, followed by the K2.5 model topping open-source charts and reaching $100 million in Annual Recurring Revenue (ARR) early this year, the team has consistently demonstrated robust engineering execution.
Very recently, on April 21st, Kimi released and open-sourced the K2.6 version, focusing on enhanced cluster processing capabilities. Merely a week later, news of K3 emerged. Moonshot's president, Zhang Yuting, has previously emphasized the team's high efficiency in resource utilization. As K3's launch approaches, the race for trillion-parameter computational supremacy is entering a critical phase.
Anthropic Expands Claude AI Coding Tools to Japan in Push for Overseas Growth
Anthropic, a prominent U.S. artificial intelligence firm, is intensifying its global outreach. On Wednesday, the company hosted a major developer gathering, "Code with Claude," in Tokyo, drawing close to 500 software engineers. This initiative seeks
India’s Software Giant Cuts Hiring, Promises No Layoffs as AI Agents Scale Up
As artificial intelligence reshapes traditional labor-intensive business models, Tata Consultancy Services (TCS), a premier Indian software outsourcing firm, has unveiled its strategic response. During Tuesday’s annual general meeting, the chairman o
China Locks AI Models During Gaokam to Block Instant Homework Help
With the 2026 Gaokao fast approaching, rumors regarding the suspension of AI tools during the exam period have ignited intense online debate. In response to public concern, major AI platforms and educational apps have clarified their stance: rather t





Home






