Model Comparison & Capabilities

Gemini CLI supports multiple Gemini models, each optimized for different workloads. Gemini 2.5 Pro is the flagship — 1M token context, strongest reasoning, best code generation quality. Gemini 2.5 Flash is the speed-optimized variant — faster responses, lower cost, shorter context, adequate quality for routine tasks. Gemini 2.0 Flash offers even faster latency at the cost of reduced capability. The model landscape evolves rapidly, but the selection principle remains constant: match the model to the task, not the task to the model.

Capability differences between models are not uniform across task types. Pro models excel at multi-step reasoning, complex debugging, architecture analysis, and code that requires understanding large dependency graphs. Flash models perform nearly as well on isolated tasks — single-file edits, formatting, documentation, simple refactoring. The performance gap widens as task complexity increases. A Pro model debugging a race condition across three services will outperform Flash significantly. Both models editing a README produce essentially identical results.

{
  "model": "gemini-2.5-pro",
  "profiles": {
    "fast": {
      "model": "gemini-2.5-flash",
      "thinkingBudget": "low"
    },
    "deep": {
      "model": "gemini-2.5-pro",
      "thinkingBudget": "high"
    },
    "review": {
      "model": "gemini-2.5-pro",
      "thinkingBudget": "auto"
    }
  }
}