Fast ~25–30 intelligence — Designed for speed without dropping to a lightweight experience. Use it for routine questions, concise research, rewriting, summaries, brainstorming, and straightforward code. Good for “Tighten this email.” · “Summarize these notes.” · “Explain this clause.” · “Fix this CSS.” Strong general-purpose reasoning; competitive with capable mainstream assistants. Context capacity Large · ~1M-class
Smart ~34–37 intelligence — The default for serious work. It spends more reasoning on ambiguity, comparison, synthesis, coding, drafting, and professional analysis while remaining responsive. Good for “Compare these contracts.” · “Analyze the tradeoffs.” · “Debug this architecture.” · “Draft a substantive memo.” Frontier-class on many reasoning tasks; around the stronger end of mainstream premium assistants. Context capacity Large · ~1M-class
Deep ~42 intelligence — For hard, multi-step problems where the quality of the reasoning matters more than getting the fastest possible answer. Good for “Stress-test this legal theory.” · “Design the system.” · “Work through this technical problem.” · “Find weaknesses in my analysis.” High frontier-class reasoning, competitive with strong premium reasoning systems. Context capacity Large · ~1M-class