🧐👉 Copilot Just Got a 1/12 Price Model—But It's From Beijing #QixNewsAI

⚡️ Quick Start: How to Try Kimi K2.7 Code

- For Copilot Subscribers: Open VS Code (v1.127.0+), Visual Studio (v17.14.6+), or JetBrains (v1.9.1-251+). In the Copilot Chat model picker, select "Kimi K2.7 Code". Available on Pro, Pro+, and Max plans.
- For Self-Hosters: Download the open weights for free on Hugging Face: https://huggingface.co/moonshotai/Kimi-K2.7-Code. Requires vLLM or SGLang. The model uses INT4 native quantization and needs roughly 240GB of storage.
- For Business/Enterprise Admins: The model is off by default. Go to Settings 》 Code, planning, and automation 》 Copilot 》 Models and set the Kimi K2.7 Code policy to "Enabled". Review your data governance requirements first.

🇨🇳 The Big News

GitHub Copilot has integrated its first open-weight model, and it's a significant one. Kimi K2.7 Code, built by Beijing-based Moonshot AI, is now generally available in the model picker. This marks a major shift, giving developers a drastically cheaper option directly inside their primary coding tool.

💰 The Price Advantage

The headline is the cost. Kimi's API is priced at roughly $0.95 per million input tokens and $4.00 per million output tokens. That makes it about 1/12th the cost of Fable 5 ($10/$50). This aggressive pricing is made possible by its Mixture-of-Experts (MoE) architecture, which activates only 32 billion of its 1 trillion total parameters per token.

⚠️ The Fine Print

- Benchmarks Are Self-Reported: As of launch, no independent third-party results exist on SWE-bench or LiveCodeBench. All performance claims come from Moonshot's own Kimi Code Bench v2.
- Performance Regressions: Community testing reveals the model's KernelBench-Hard MoE kernel score dropped from 0.222 (K2.6) to 0.157.
- Licensing Trap: The model uses a Modified MIT License. If your app has over 100 million monthly active users or $20M+ monthly revenue, you must display attribution. Cursor was previously caught violating this with K2.5.
- Legal Jurisdiction: Moonshot AI is subject to China's National Intelligence Law, which can compel companies to cooperate with state intelligence efforts. While inference runs on Microsoft Azure in the US, the legal obligation applies to Moonshot as a company.

🧠 Our Verdict

For individual developers, this is a no-brainer to test. The cost savings are real, and you can switch in 5 minutes. For enterprises, the decision is more complex. The combination of unverified benchmarks, a restrictive license, and serious jurisdictional questions means you should complete a full legal and security review before enabling it for your team. The model is off by default for a reason.

#Kimi_K2.7_Code #GitHub_Copilot #Open_Weight_Model #Moonshot_AI #AI_Coding_Tools #QixNewsAI #Shorts

Видео 🧐👉 Copilot Just Got a 1/12 Price Model—But It's From Beijing #QixNewsAI канала QixNews