Model Selection

Choose the right Gemini model for your task.

Available Models

gemini-3.1-pro-preview (Default)

• Best for: Complex analysis, large codebases, architectural reviews
• Context: 1M tokens
• Use when: You need the strongest reasoning, code review, or architectural analysis

gemini-3-flash-preview (Fallback)

• Best for: Quick responses, routine tasks, fast iteration
• Context: 1M tokens
• Use when: Speed matters more than depth, or Pro quota is exhausted

Legacy Models

Older models (gemini-2.5-pro, gemini-2.5-flash) are still supported by the Gemini CLI. You can explicitly request them via the model parameter, but they are no longer the default.

Setting Models

You need use natural language: "...using gemini flash"

You can also append with '-m' or ask specifically with

Model Comparison

Model	Speed	Context	Best Use Case
gemini-3.1-pro-preview	Slower	1M tokens	Complex reasoning, big ideas
gemini-3-flash-preview	Fast	1M tokens	Quick, specific changes

Cost Optimization

1. Start with Pro (default) for most tasks — best quality
2. Use Flash when you need faster responses or hit Pro quota limits
3. The tool automatically falls back to Flash when Pro quota is exceeded

Token Limits

• Pro: ~1 million tokens (~250k lines of code)
• Flash: ~1 million tokens (~250k lines of code)

Recommendations

• Code Review: Pro (default)
• Architecture Analysis: Pro
• Quick Fixes: Flash
• Documentation: Flash
• Security Audit: Pro