Reasoning Effort

Kavia allows you to set the reasoning depth for each task. This determines how much analysis the AI performs:

• Minimal – Fastest and most efficient, best for simple edits or quick tasks.
• Low – Slightly deeper analysis, good for straightforward builds or small fixes.
• Medium – Balanced option, recommended for most workflows and troubleshooting.
• High – Most detailed reasoning, ideal for complex coding, architecture, or advanced problem-solving.

Model Selection

Kavia supports multiple AI models, giving flexibility based on your needs:

• GPT-5.2 – Latest generation model with the highest accuracy, strongest reasoning, and advanced coding capabilities. Best for complex logic, large systems, and mission-critical workflows.
• GPT-5.1 – Powerful general-purpose model optimized for balanced reasoning, coding, and long-context tasks. Ideal for end-to-end feature development and architectural discussions.
• Gemini-3-Pro - Next-generation Gemini model designed for advanced reasoning, large-scale systems, and high-performance AI applications.
• Gemini-2.5-Pro – Strong multi-modal model capable of handling text, images, and broad reasoning tasks. Well suited for AI-rich workflows and complex content generation.
• Claude 4.0 - Excellent for natural language understanding, structured writing, deep analysis, and handling complex or ambiguous instructions.
• Claude 3.5 - Faster and lighter Claude model with solid reasoning, ideal for text-heavy tasks and quick iterations.
• GPT-4.1 – Reliable for coding and documentation with lower resource use.
• Azure GPT-4.1 – Enterprise-grade GPT-4.1 with Microsoft-hosted compliance/security.
• O3 – Lightweight and fast, best for simple or repetitive tasks.