GitHub put a Chinese open model in Copilot. The thread was too busy leaving over the price hike to notice
Kimi K2.7 Code landing in GitHub Copilot is a real milestone — a way to run a Chinese open-weight model from a trusted host. But the comments turned into a referendum on Copilot's June pricing change and the customers it drove out.
The launch itself is clean: Kimi K2.7 is selectable across VS Code, JetBrains, Xcode and the rest, hosted on Azure, billed usage-based. At Moonshot's rates, one reader clocked it at roughly GPT-5.4-mini territory.
Then the real conversation took over. A Copilot CLI superfan described his whole department burning the monthly budget every two to three days after June and defecting to Claude Code, guessing Copilot 'lost 50% or more of their customer base last month.' A small team hit an undocumented 10-license minimum and left for Cursor.
The milestone, stated plainly
GitHub announced that Kimi K2.7 Code is generally available in Copilot — rolling out first to Pro, Pro+ and Max, with Business and Enterprise to follow (off by default; admins must enable it). It's hosted by GitHub on Microsoft Azure, billed usage-based at provider list pricing, and selectable across a wide surface: VS Code, Visual Studio, the Copilot CLI and cloud agent, JetBrains, Xcode and Eclipse. The genuine significance, as one reader framed it, is a trusted-provider path to run a Chinese open-weight model. Its pricing simply mirrors Moonshot's own list rates:
| Kimi K2.7 (per Moonshot list) | Rate |
|---|---|
| Input | $0.95 |
| Cache hit | $0.19 |
| Output | $4.00 |
which one reader who priced it out put "at roughly the price of GPT 5.4 mini."
The comments were a referendum on Copilot's June invoice
The dominant story wasn't Kimi at all — it was churn:
"I am a huge fan of Copilot CLI… Unfortunately the June pricing change for Copilot forced me personally as well as my entire department at work to switch to Claude Code… in June we started chewing through the monthly budget every 2-3 days… I wouldn't be surprised if Copilot lost 50% or more of their customer base last month." — a Copilot CLI user describing his department's forced migration
And a smaller team hit a wall that isn't in any pricing page:
"you will waste hours navigating their billing settings… I finally got an email from them saying that "Copilot Business is available for teams purchasing 10 or more licenses". This is undocumented but other people are reporting the same… We're sticking with Cursor for now, using Kimi as our daily driver." — a small-team buyer who hit an undocumented 10-license minimum
So the headline event — a Chinese frontier-ish model, one click away inside the world's default code host — arrives as a footnote to its own launch. The model isn't the thing customers are reacting to. The bill is. GitHub opened the gate to exactly the cheaper models people wanted, in the same quarter it gave a chunk of them a reason to walk through someone else's.
Get breakdowns like this every day
The top Hacker News stories, unpacked — by email, Telegram, or RSS.
Cursor's benchmark says its cheap model rivals the frontier. A rival benchmark says it scores 16
CursorBench 3.1 shows Cursor's own Composer 2.5 landing near the top coding models at one-tenth the price. The thread pulled up an independent benchmark where the same model scores 16 to GPT-5.5's 64.