Fast ~9–13 intelligence — Best for quick factual questions, rewriting, summaries, brainstorming, short calculations, and routine chat. Examples “Rewrite this email.” · “Summarize this article.” · “Give me five ideas.” Typical context ~128K tokens
Smart ~20–28 intelligence — Best for substantive analysis, comparing options, drafting, coding help, research synthesis, and most professional work. Examples “Compare these contracts.” · “Debug this function.” · “Analyze these alternatives.” Typical context ~128K–256K tokens
Deep ~28–34 intelligence — Best when the problem is difficult, ambiguous, multi-step, or worth spending more reasoning on before answering. Examples “Stress-test this legal argument.” · “Design the architecture.” · “Work through this complex problem.” Typical context ~128K–256K tokens