Moonshot AI, the two-year-old Beijing lab backed by Alibaba, just released Kimi K2 Thinking, an open-source reasoning model it says can match or beat OpenAI’s GPT-5 and Anthropic’s Claude 4.5 Sonnet—while costing just $4.6 million to train. (CNBC, ZDNet) The reported bill, sourced to an anonymous insider quoted by CNBC and amplified by open-source communities, is roughly equivalent to the fully loaded cost of a small Silicon Valley engineering pod. (CNBC, Reddit)
Built on top of the July Kimi K2 release, the new model leans on the same DeepSeek-derived architecture but layers heavier tool use and autonomous planning. Benchmarks reported by observers show Kimi K2 Thinking hitting GPT-5-class scores on BrowseComp and Humanity’s Last Exam, positioning it as the first fully open model to challenge the Western frontier. (Interconnects, X/Twitter)
Moonshot AI was founded in March 2023 by 32-year-old Yang Zhilin, a Tsinghua-trained computer scientist who later earned his PhD at Carnegie Mellon before rotations through Google Brain and Meta AI. (Antoine Buteau, Yahoo Finance, ChinaTalk) Yang co-led the research behind Transformer-XL and says the startup’s name nods to Pink Floyd’s “Dark Side of the Moon,” signaling an ambition to explore the edge of AGI research. (BytePlus, BytePlus)
Even Kimi’s brand is personal: the chatbot takes its name from Yang’s own English nickname, underscoring how closely his identity and the product are entwined. (BytePlus)
Alibaba disclosed in its 2024 annual report that it bought a 36% stake in Moonshot for roughly $800 million, valuing the startup around $2.2 billion before newer rounds reportedly lifted the figure closer to $3.3 billion. (Yahoo Finance, SCMP, Pandaily) The e-commerce giant has now backed all four of China’s so-called AI Tigers—Baichuan, Zhipu, MiniMax, and Moonshot—giving it outsized sway over the country’s next generation of frontier labs. (SCMP) Tencent has reportedly joined the cap table in recent financings as well. (Yahoo Finance)
Analysts often bracket Moonshot with DeepSeek because both shocked Silicon Valley with unexpectedly strong, low-cost releases, yet their philosophies diverge. (Thoughtworks, KR Asia) DeepSeek’s R1 leaned into chain-of-thought transparency, staffing primarily with domestically trained engineers and showcasing methodical reasoning. (ChinaTalk, KR Asia, Appy Pie Automate)
Moonshot instead recruits from both U.S. and Chinese elite schools and prioritizes agentic behavior over visible deliberation. Kimi K2 Thinking is optimized to take action, not just reason, chaining 200 to 300 tool calls through browsers, spreadsheets, and even 3D software with minimal human oversight. (KR Asia, ChinaTalk, Thoughtworks, LinkedIn, Reddit) Internally, the team concluded that iterating on DeepSeek-V3’s layout beat reinventing the stack, so it poured effort into parameter scaling, optimizer tweaks, and inference cost controls. (Recode China AI)
Moonshot first broke out in October 2023 when Kimi Chat processed 200,000 Chinese characters—novel-length context that rivals struggled to touch—cementing “long context” as its calling card. (Pandaily, ChinaTalk, BytePlus) By July 2025, the standard Kimi K2 release prompted Nature to call it “another DeepSeek moment” after it became the fastest-downloaded model on Hugging Face while staying open-weight and cheap to run. (Nature, KR Asia)
Kimi K2 Thinking vaults Moonshot from a long-context specialist to a frontier competitor. The model posts 44.9% on Humanity’s Last Exam with tools, compared with GPT-5’s 41.7%, and 60.2% on BrowseComp versus GPT-5’s 54.9% and Claude 4.5’s 24.1%. (36Kr, Cybernews, LinkedIn) If the $4.6 million training tab holds up, it signals a dramatic efficiency gap powered by architecture reuse, data curation, and cheaper Chinese compute. Industry watchers still peg U.S. frontier training runs at hundreds of millions to billions of dollars. (Reddit, Gelonghui, Ifeng Tech, ZDNet)
Open-weight access plus near-frontier scores is forcing U.S. incumbents to reassess their moats. As one analyst noted after testing the agentic workflows, the old assumption that Chinese labs trailed by years now looks dangerously outdated. (ZDNet, LinkedIn)