JetBrains launched Junie Local, a free coding agent that runs fully on your machine.
It matches Claude Sonnet 4.5 in internal tests.
It currently works only on high-end Macs, but broader support is coming.
What happened
JetBrains announced Junie Local, a coding agent that runs entirely on a local machine. It is free to download and use, with no API or model usage fees, and no code leaves the machine. Users switch from cloud Junie with the command "/local" and can return via "/model".
Why it matters
Junie Local is positioned as the local version of JetBrains' cloud-based agent Junie. In internal tests, it matched Claude Sonnet 4.5 (with a 10,000-token reasoning limit) and was slightly edged out by GPT-5 set to "medium effort". This makes free, cost-sensitive tasks like multi-file refactoring or fixing low-priority test coverage more accessible.
What to watch
Junie Local currently runs on Macs with the M5 processor and 64GB memory, which JetBrains calls a starting point. Prototypes run on PCs with RTX5090 GPUs and DGX Spark, and support for other 24GB-memory GPUs is under consideration.
Ask the AI about this article →
JetBrains' move to offer a local, free coding agent reflects a broader push in developer tools toward on-device AI, where code stays private and running costs drop to zero. The company already offered a cloud-based Junie, so Junie Local extends that product line rather than replacing it. Users can switch between the two with simple commands, and existing configuration files like AGENTS.md remain valid, lowering the barrier to trying the local version.
The choice of the Qwen3.6-27B model, quantized to 4 bits, suggests a focus on balancing capability with the hardware constraints of a local machine. Internal tests showing parity with Claude Sonnet 4.5 position Junie Local as a credible free alternative for routine coding tasks, though GPT-5 still holds a slight edge under "medium effort" settings. The current hardware requirement—M5 Macs with 64GB memory—limits the audience, but JetBrains' stated plans to support RTX5090 PCs and DGX Spark, plus other 24GB GPUs, point to a deliberate expansion path.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
A Pew Research Center survey of 3,488 U.S
Caterpillar is applying lessons from its mining automation business to broader AI deployment, including a voic…
A study by two independent AI researchers, done as part of the MATS research program, tested Anthropic's Claud…
A randomized experiment with 1,053 freshmen at Bocconi University found that GPT-4o helped students earn signi…
OpenAI and METR published final technical reports on the July Hugging Face breach
Anthropic is effectively cutting weekly usage limits for Claude Code by 17 percent
