Model selection

LLM providers

Junie supports four types of LLM providers:

• Junie — models accessed through a JetBrains AI subscription, by means of a Junie login or a JetBrains AI API key. No additional configuration is required.

• BYOK (Bring Your Own Key) — models accessed using your own API key from a third-party provider such as OpenAI, Anthropic, Google, xAI, or OpenRouter. See BYOK for setup instructions.

• Custom — models defined in a custom profile JSON file. See Custom LLM models for details.

• Proxy — models accessed through a custom proxy endpoint configured in config.json. See Custom proxies for setup instructions.

When a model is available through multiple providers, Junie prefers the Junie provider by default. Use --provider to override this behavior.

How Junie uses models internally

The model you select (or the Default) is your primary model — Junie uses it for the main reasoning and code generation work.

However, Junie also uses a separate model for certain internal tasks such as:

• Summarizing context

• Classifying and routing tasks

• Extracting information for memory

• Filtering capabilities

• Other helper tasks

These internal tasks don't need the full power of the primary model, so Junie automatically picks an alternative from the same provider. Note that these models are not always smaller — for example, routing may use GPT-4.1 by default. If your primary model is from Anthropic, the helper model will typically be Claude Haiku; if it's from Google, it will be Gemini Flash.

This applies even when using BYOK. If you bring your own API key, Junie will still use a separate model from the same provider for these internal tasks. This means you may see API calls to a model you didn't explicitly select — that's expected and helps keep costs down and responses fast. For more details on configuring custom LLM providers, see Custom LLM Models.

Junie CLI also supports model aliases. Each model alias ID always points to the latest supported version of that model.