Junie Lite
Junie Lite is a free baseline model tier for Junie, JetBrains' coding agent. It is not a trial, not a limited-time sample, and it does not expire or count down.
The model currently powering Junie Lite is Qwen 3.6 27B. This is an implementation detail, not a commitment: JetBrains may change the underlying model over time without notice. What is fixed is the tier itself — a free baseline, always available.
Quick start
Sign in with a valid JetBrains account in Junie — via JetBrains Air in your IDE, or Junie CLI.
Junie Lite access is granted automatically on your first request to the model. No opt-in, no separate signup, and no card required.
To select Junie Lite manually instead of letting Junie route to it, run
/modelin Junie CLI. See Selecting a model manually.
The rest of this page explains each of these steps in detail.
Selecting a model manually
In JetBrains Air or Junie CLI, open the model selector, or run /model, to see the available models and select Junie Lite directly.
For model selection in headless and scripted runs, see Model selection. For the /model command reference, see Slash commands.
Which model powers Junie Lite
Junie Lite currently runs on Qwen 3.6 27B. JetBrains may swap the underlying model over time — for example, to a different Qwen version, or to another JetBrains or third-party model — without a change to the Junie Lite tier itself or advance notice.
Do not build workflows around the specific model name; build around the tier.
Usage limits and quotas
Junie Lite does not publish a quota number, because the underlying model can change. Virtually unlimited describes normal use, not literally uncapped requests — the underlying model has its own capacity ceiling.
If a model's own capacity limit is reached, requests stop rather than falling back to another model — there is no downgrade path from Junie Lite. Junie Lite also does not publish a tool-call count limit.
Data handling and compliance
Inference for Junie Lite runs exclusively on JetBrains infrastructure. No customer data, prompts, outputs, or telemetry are transferred to the model provider or disclosed to any subprocessor. There are no side agreements, data-sharing arrangements, or model-improvement flows with the model provider — your code is not used to train the underlying model.
Availability
Junie Lite is included automatically with every JetBrains account: free accounts, and active AI Pro and AI Ultimate license holders. It is available in JetBrains Air in your IDE and in Junie CLI.
FAQ
Is Junie Lite a trial?
No. It is a standing free tier, not a trial — it does not expire or count down.
What model powers Junie Lite?
Currently Qwen 3.6 27B. This may change without notice; the tier name stays the same.
Can I select Junie Lite manually?
Yes. In Junie CLI, run /model. In JetBrains Air, use the model selector. See Selecting a model manually.
Is my code used to train the model?
No. Inference runs exclusively on JetBrains infrastructure; no code, prompts, or outputs are shared with the model provider or used to improve their models.
What happens if I hit the capacity limit?
If a model's own capacity limit is reached, requests stop rather than falling back to another model — there is no downgrade path from Junie Lite.
Why doesn't JetBrains publish a specific quota number for Junie Lite?
Because the underlying model can change over time, and a published number would tie the tier to whatever model happens to power it today. Junie Lite is described as virtually unlimited for normal use rather than as a fixed allowance.