Learn
What is reasoning effort?
Last reviewed July 16, 2026
Reasoning effort is the control on reasoning models that sets how much hidden thinking a model does before it answers: low effort replies fast and cheap, high effort spends more reasoning tokens on hard problems. The thinking is billed as output tokens, so effort is a per-request trade between answer quality, latency, and cost, set by the caller rather than fixed per model.
What effort actually changes
Reasoning models generate internal chain-of-thought tokens before the visible answer. Effort tiers (none, low, medium, high on most APIs) budget that thinking. On hard benchmark suites the same model can score meaningfully higher at high effort, at multiples of the latency and token spend; on routine questions extra effort mostly buys slower, costlier answers of equal quality.
Choosing a tier
Match effort to task difficulty, not habit: extraction, formatting, and lookups run at none or low; multi-step math, hard debugging, and planning earn high. Because thinking bills as output tokens (the expensive kind), defaulting everything to maximum effort is the most common silent cost multiplier in reasoning-model use.
idapt exposes reasoning effort per message on models that support it and benchmarks reasoning models per effort tier, so model pages show what each level of thinking actually buys.
Frequently asked
See it in practice
The fastest way to understand it is to use it. Start free, no credit card.
Try idapt free