Kimi (chatbot)
| Kimi | |
|---|---|
Screenshot Screenshot of an example of a Kimi K2 answer describing Wikipedia | |
| Developer | Moonshot AI |
| Release | October 2023 |
| Stable release | Kimi K3 / 16 July 2026 |
| Platform | |
| Type | Chatbot Large language model |
| License | Proprietary MIT License (Kimi-VL, Kimi-Dev) Modified MIT License (Kimi K2) |
| Website | kimi.com platform (API platform) |
Kimi is an artificial intelligence (AI) chatbot and series of large language models developed by Chinese company Moonshot AI. Its first version, released in 2023, was known for supporting up to 128,000 tokens of context.[1] The company released the open-weights Kimi K2 in July 2025,[2][3] and Kimi K3 in July 2026.
History
[edit]Moonshot AI was founded in March 2023 in China. In October 2023, the company officially released the Kimi chatbot and began closed-beta testing.[4]
On 16 November 2023, Kimi was released to the general public based on the Moonshot model. The first version of Kimi supported lossless context of 128,000 tokens,[1] making it the first AI model that was capable of accepting contexts of this size.[5]
In March 2024, Moonshot AI began closed beta testing of an updated version of Kimi with a 2 million character context window.[6]
In July 2024, Kimi's "context caching" feature entered public beta.[7]
On 11 October 2024, Kimi Explore Edition, featuring AI-powered autonomous search, went live globally;[8] monthly active users have since exceeded 36 million.[9][10]
In November 2024, Kimi began internal testing of an AI video generation model.[11]
On 20 January 2025, Kimi K1.5 was released. Moonshot AI claimed it matched the performance of OpenAI o1 in mathematics, coding, and multimodal reasoning capabilities.[12]
In April 2025, Kimi-VL, an open source 16 billion parameter mixture of experts (MoE) large language model with 3 billion active parameters, was released. In June, a reasoning model named Kimi-VL-Thinking was also released.[13][14]
In June 2025, Kimi-Dev, a 72B parameter coding-focused model based on Qwen2.5-72B, was released.[15][16] Also in June, Moonshot AI released Kimi-Researcher, an autonomous AI research agent available through Kimi's website and app.[17][18]
In July 2025, Moonshot AI released Kimi K2, a 1 trillion parameter mixture of experts large language model with 32 billion active parameters which was open sourced under a modified MIT license.[3][2] On 9 September 2025, Moonshot AI released an updated version of K2, Kimi-K2-Instruct-0905, which improved its performance in coding tasks and increased its context window from 128K tokens to 256K tokens.[19][20]
In September 2025, Moonshot AI added an agentic AI feature to Kimi known as "OK Computer" (likely named after the Radiohead album of the same name). It is capable of creating multi-page websites and editable slides from simple user prompts and can process up to 1 million rows of input data at once, and output text, audio, images, and video.[21][22]
In October 2025, Moonshot AI released Kimi Linear, a 48 billion parameter MoE model with 3 billion active parameters. It uses a more efficient attention method known as Kimi Delta Attention (KDA), which reduces memory usage and improves generation speed at longer context window sizes.[23]
In January 2026, Moonshot AI released Kimi K2.5, a 1 trillion parameter MoE model with 32 billion active parameters. The model is multimodal in vision and language understanding with advanced agentic capabilities, instant and thinking modes, as well as conversational and agentic paradigms.[24] Moonshot AI released Kimi K2.6 in April 2026,[25][26][27] followed by Kimi K2.7 Code in June 2026.[28][29]
Moonshot AI released its Kimi K3 model, on 16 July 2026, with 2.8 trillion parameters. Upon release, Kimi K3 debuted at No. 3 on the Artificial Analysis AI leaderboard, behind Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol, but outperformed its competitors on Arena.ai's front-end web development benchmark. Moonshot AI announced that it would release the weights for Kimi K3 by 27 July.[30]
Model versions
[edit]| Version | Release date | Description |
|---|---|---|
| K2 Thinking | November 2025 | Reasoning model |
| K2.5 | January 2026 | |
| K2.6 | April 2026 | [25][26][27] |
| K2.7 Code | June 2026 | [28][29] |
| K3 | July 2026 |
Pricing
[edit]The Kimi apps are free to use with general rate limits. However, there are four subscription plans known as "Moderato", "Allegro", "Allegretto", and "Vivace" (all named after tempo markings). The plans offer higher usage of the K2 model, access to K2 Turbo, a version of the model that runs on faster hardware, extended access to Kimi Researcher and OK Computer, and faster Kimi slides generation.[31]
See also
[edit]References
[edit]- 1 2 Xu Ziming. "Domestic large models "roll out": Kimi's ultra-long context goes viral; Wondershare "Tianmu" focuses on audio & video". caijing.chinadaily.com.cn. Archived from the original on 2025-01-20. Retrieved 2024-12-29.
- 1 2 Cheng, Evelyn (2025-07-14). "Alibaba-backed Moonshot releases new Kimi AI model that beats ChatGPT, Claude in coding — and it costs less". CNBC. Retrieved 2025-09-09.
- 1 2 "Kimi K2: Open Agentic Intelligence". moonshotai.github.io. Retrieved 2025-09-09.
- ↑ "Moonshot AI's large-model service "Kimi Chat" starts closed beta, in deep co-operation with Volcengine". company.cnstock.com. Archived from the original on 2023-10-17. Retrieved 2024-12-29.
- ↑ Cao Jing. "Supporting 200 000-character input, Moonshot AI opens the "long-text" era for hundred-billion-parameter models". cn.chinadaily.com.cn. Archived from the original on 2025-01-21. Retrieved 2024-12-29.
- ↑ "Moonshot's Kimi intelligent assistant supporting 2-million-character context starts internal testing". news.cnstock.com. Retrieved 2024-12-29.
- ↑ Tencent (2024-07-01). "Moonshot Kimi open platform launches public beta of "context caching"". news.qq.com (in Chinese (China)). Archived from the original on 2025-01-20. Retrieved 2024-12-29.
- ↑ Tencent (2024-10-12). "One search can deeply read 500+ pages: Kimi Explore Edition arrives—will AI out-search humans?". news.qq.com (in Chinese (China)). Archived from the original on 2025-01-20. Retrieved 2024-12-29.
- ↑ "AI weekly: Yang Zhilin says Kimi MAU exceeds 36 million; Robin Li: large-model hallucination basically eliminated". www.yicai.com. Archived from the original on 2025-01-20. Retrieved 2024-12-29.
- ↑ "AI assistant Kimi signs with Chengdu Island of Scientific Innovation". www.cdrb.com.cn. Archived from the original on 2025-01-20. Retrieved 2024-12-29.
- ↑ Guancha (2024-11-28). "Kimi testing AI video generation? Moonshot: currently in grey-box testing". t.cj.sina.cn. Archived from the original on 2025-01-20. Retrieved 2024-12-29.
- ↑ "Chinese AI Firms Debut New LLMs to Rival OpenAI's Powerful O1 in Math and Coding". www.yicaiglobal.com. Retrieved 2025-09-09.
- ↑ MoonshotAI/Kimi-VL, Moonshot AI, 2025-09-09, retrieved 2025-09-09
- ↑ Kemper, Jonathan (2025-04-27). "Moonshot AI's open-source Kimi-VL tackles text, images and video with just 2.8 billion parameters". THE DECODER. Retrieved 2025-09-09.
- ↑ MoonshotAI/Kimi-Dev, Moonshot AI, 2025-09-08, retrieved 2025-09-09
- ↑ "China's AI sprint: MiniMax, Moonshot lock horns in multimodal model race". DIGITIMES. 2025-06-24. Retrieved 2025-09-09.
- ↑ "Kimi-Researcher: End-to-End RL Training for Emerging Agentic Capabilities". moonshotai.github.io. Retrieved 2025-09-09.
- ↑ "Kimi from Moonshot Dark Side Launches Kimi-Researcher Deep Research Agent and Starts Internal Testing". AIbase. Retrieved 2025-09-09.
- ↑ "moonshotai/Kimi-K2-Instruct-0905 · Hugging Face". huggingface.co. 2025-09-05. Retrieved 2025-09-09.
- ↑ "Moonshot AI's updated Kimi model offers expanded context window, improved coding". South China Morning Post. 2025-09-04. Retrieved 2025-09-09.
- ↑ "Moonshot AI's Kimi chatbot offers 'agent mode' for creating websites, slides". South China Morning Post. 2025-09-25. Retrieved 2025-10-31.
- ↑ "Kimi launches new Agent mode OK Computer | Bitget News". Bitget. Retrieved 2025-10-31.
- ↑ "moonshotai/Kimi-Linear-48B-A3B-Instruct · Hugging Face". huggingface.co. 2025-10-31. Retrieved 2025-10-31.
- ↑ "Kimi K2.5: Visual Agentic Intelligence". www.kimi.com. Retrieved 2026-01-27.
- 1 2 Gewirtz, David (21 April 2026). "Moonshot AI's new Kimi K2.6 swarms your complex tasks with 1,000 collaborating agents". ZDNET. Retrieved 17 June 2026.
- 1 2 Li, Yutong (22 April 2026). Du, Yu (ed.). "13小时编码、5天自主运行!Kimi K2.6开源"硬刚"闭源巨头,长程战力能否撕开AGI工业化新赛道?" [13 hours of coding, 5 days of autonomous operation! Kimi K2.6 open-source "goes head-on" against closed-source giants. Can its long-term capabilities carve out a new path for AGI industrialization?]. National Business Daily (in Chinese). Retrieved 17 June 2026.
- 1 2 "Kimi K2.6 Tech Blog: Advancing Open-Source Coding". www.kimi.com. Retrieved 2026-04-20.
- 1 2 "The Chinese-made AI model 'Kimi K2.7 Code' has been released as an open model, boasting the best coding performance in the Kimi series and low token consumption". Gigazine. 15 June 2026. Retrieved 17 June 2026.
- 1 2 "Kimi K2.7 Code: Open-Source Agentic Coding Model". Kimi. 16 June 2026. Retrieved 17 June 2026.
- ↑ Yildirim, Ece (16 July 2026). "China Just Dropped Another Bomb on America's Frontier AI Companies". Gizmodo. Retrieved 16 July 2026.
- ↑ "Kimi - An AI assistant capable of reasoning, analysis, and deep thinking. (translated from Chinese)". www.kimi.com. Retrieved 2025-10-31.