Kimi’s New Funding Round

Kimi is set to complete a new funding round of $2 billion, with a post-money valuation exceeding $20 billion. This round is led by Meituan Longzhu, with participation from China Mobile, CPE Yuanfeng, and several existing shareholders.

To date, Kimi’s total funding has surpassed 37.6 billion RMB, making it the most funded startup in China’s large model sector. Its annual ARR revenue has also grown from over $100 million in early March to over $200 million in April, driven primarily by paid subscriptions and API calls.

Breakthrough with K2.6 Model

Two weeks ago, Kimi released its K2.6 model, boasting long-range coding capabilities with a 1T parameter MoE architecture and 32B activations, allowing for continuous coding for 13 hours and enabling an agent to operate autonomously for five days. This achievement has reclaimed the top position among open-source models globally.

K2.6 has demonstrated performance on par with or exceeding that of top proprietary models like GPT-5.4, Claude Opus 4.6, and Gemini 3.1 Pro in several authoritative benchmarks. It integrates coding capabilities with deep visual understanding, delivering highly creative professional-grade web applications, which could revolutionize front-end development and product prototype design.

The architecture driven by K2.6, known as the “Agent cluster,” has undergone significant upgrades, supporting 300 sub-agents to collaboratively complete 4,000 steps. Compared to the previous K2.5 generation, both task completion and delivery quality have seen substantial improvements.

Innovations in Hardware Compatibility

Both Kimi and DeepSeek are exploring adaptations for domestic chips. DeepSeek V4 has deep compatibility with Huawei Ascend chips, while Kimi’s latest paper, “Prefill-as-a-Service,” proposes a heterogeneous hardware inference framework across data centers, allowing different types of domestic chips to handle the Prefill and Decode stages. This approach has shown a 54% increase in throughput and a 64% reduction in first token latency, facilitating the integration of domestic chips into the large model inference chain.

According to the Artificial Analysis intelligent index open-source model ranking, all top five open-source models globally are Chinese, with Kimi K2.6 ranking first. Developers have reported that K2.6’s programming capabilities even surpass those of DeepSeek V4, establishing Kimi as one of the strongest players in the global open-source arena.

Shifting Dynamics in the AI Landscape

The recent developments mark a significant shift in the narrative surrounding Chinese entrepreneurs, who have traditionally looked to Silicon Valley as the benchmark for technology. China’s AI sector is carving its own path, leveraging domestic chips to run Chinese models for global open-source contributions.

Data supports this trend, with OpenRouter reporting that in Q1, the weekly token call volume for Chinese open-source large models exceeded 60%, with Kimi and DeepSeek leading the charge. The global discourse on AI foundational technologies is gradually shifting.

The Rise of Young Innovators

Yang Zhilin, born in 1992 and the youngest entrepreneur at a recent meeting hosted by the Prime Minister, represents a new wave of innovation in AI. Just over a year ago, DeepSeek’s founder Liang Wenfeng was in a similar position, having just gained global recognition, while Kimi faced growth challenges.

In a year, the landscape has changed dramatically: DeepSeek has launched its much-anticipated V4 model, while Kimi has made a remarkable comeback through foundational architecture innovation, gaining recognition in Silicon Valley.

Historically, Kimi and DeepSeek have often been compared, with their releases frequently coinciding. This time, the K2.6 and DeepSeek V4 releases were only days apart. While DeepSeek initially appeared to have the upper hand, Kimi’s focus on foundational architecture has begun to pay off, establishing both companies as pivotal players in the global AI ecosystem.

Investment Trends and Future Outlook

Investors are increasingly optimistic about Kimi, which has raised over $3.9 billion in less than half a year, with its latest valuation skyrocketing from approximately $4.3 billion last November to over $20 billion now. Kimi’s rapid fundraising is unprecedented in such a short timeframe.

There are indications that Kimi is considering an IPO in Hong Kong, a significant shift from its previous stance of not rushing to go public. The recent surge in AI stock prices has created a sense of urgency among investors to secure shares before the company goes public.

Kimi has advised that all financing activities are directly managed by the company, warning of risks associated with unofficial investment claims.

In this fast-paced environment, the race to secure Kimi’s shares is intense, with major players like Meituan Longzhu investing over $200 million in this round alone. The dynamics of investment are evolving, with industry players potentially becoming Kimi’s primary enterprise clients, providing computational power and channel support.

As the AI sector experiences explosive growth, the urgency for investors to engage with Kimi before its public offering has never been greater. The phrase “AI one day, human society one year” captures the essence of this era, where expectations and challenges coexist.