Cursor, Kimi & the Open Source Imperative
Cursor запустил Composer 2 для более чем миллиона ежедневных пользователей, и вскоре выяснилось, что его флагманская модель построена на китайской открытой модели Moonshot AI Kimi K2.5. Модель Cursor близка к state-of-the-art при цене в восемь раз ниже, а сам редактор основан на open-source VS Code — компания с капитализацией $50 млрд выросла на открытых фундаментах. Американские открытые frontier-модели в среднем устаревают на 8 месяцев, тогда как китайские — всего на 7 недель; Qwen обогнал Llama по загрузкам, достигнув 700 млн на Hugging Face. Meta в 2025 году ушла от open-source, а доля китайских открытых моделей в мировом использовании выросла с 1,2% до почти 30%. Однако NIST обнаружил, что китайские модели в 12 раз уязвимее к agent hijacking, и Microsoft, News Corp и госведомства запретили их. В ответ NVIDIA выделяет $26 млрд на open-source через Nemotron Coalition, а OLMo 3 от Allen Institute соответствует Qwen 3 на математике при в 6 раз меньшем объёме данных.
Last week, Cursor launched Composer 2 to over one million daily active users.1 Within hours, a developer discovered Cursor had built its flagship model on top of Moonshot AI’s Kimi K2.5, a Chinese open-source model.2
Moonshot AI’s response? “This is the open model ecosystem we love to support.”3
Cursor’s model is at near parity with state-of-the-art at one-eighth the price.4 It’s also no coincidence the editor powering Cursor is open-source, VS Code.
$50B in market cap on open-source foundations. Open source empowers startups to compete with incumbents.
It’s not easy to replicate Cursor’s innovation on US models. American open-source frontier models average 8 months old. Chinese open-source models average 7 weeks. That’s a 5x age gap. Cursor chose Kimi K2.5 (8 weeks old) over GPT-OSS (8 months old) for good reason : in AI, eight months is three generations of models.
Meta, formerly America’s open-source champion with Llama, pivoted to closed-source development in 2025.5 Chinese open-source models grew from 1.2% of global AI usage in late 2024 to nearly 30% by the end of 2025.6 Qwen overtook Llama in cumulative downloads by October 2025, reaching 700 million downloads on Hugging Face.7
But commercializing Chinese models in the US carries risks : NIST found Chinese models 12x more susceptible to agent hijacking attacks,8 & companies like Microsoft & News Corp have banned their use entirely.9 Many government agencies have followed suit.
Meanwhile, the American open-source response is taking shape. NVIDIA announced a $26 billion commitment over five years to open-source AI through its Nemotron Coalition.10 Google, OpenAI, & the Allen Institute are building alternatives. OLMo 3 matches Qwen 3 on math benchmarks with 6x less training data.111213
Cursor’s choice wasn’t ideological. It was practical. When the best open-source option is Chinese, that’s what a $50 billion company will use.
Open source is how startups compete with giants. The next Cursor will be built on the best open-source foundation available. The question is whether that foundation will be American.
TechCrunch : Cursor admits its new coding model was built on top of Moonshot AI’s Kimi ↩︎
The Decoder : Cursor quietly built its new coding model on top of Chinese open-source Kimi K2.5 ↩︎
Composio : Kimi K2.5 vs. Opus 4.5 pricing comparison - Kimi K2.5 costs $0.60/M input vs Claude Opus $5.00/M input (8.3x cheaper) ↩︎
Bloomberg : Inside Meta’s Pivot From Open Source to Money-Making AI Model ↩︎
OpenRouter : State of AI 2025 - 100T Token LLM Usage Study ↩︎
Xinhua : Alibaba’s Qwen leads global open-source AI community with 700 million downloads ↩︎
NIST : CAISI Evaluation of DeepSeek AI Models Finds Shortcomings and Risks ↩︎
TechCrunch : DeepSeek - The countries and agencies that have banned the AI company’s tech ↩︎
Allen Institute for AI : OLMo 3 - Charting a path through the model flow ↩︎