Diminishing Returns on Reasoning Model Effort Levels vs Cost
draginol · x · 2026-07-04
Empirical data on the cost of reasoning effort levels reveals a stark diminishing return: while token consumption scales drastically (Low=1×, Medium=2×, High=4×, Very High=8×, Maximum=16×), the corresponding capability gains plateau quickly (approx. 1, 2, 2.5, 2.75, 2.87). This highlights that the cost-effectiveness of high effort drops off rapidly.
More from Models
- Kimi K3.1 Pro tops MathArena open models, closing frontier gap to about 6 months — teortaxesTex · 2026-10-26
- Ruff creator Charlie Marsh: frontier models find infinite bugs in any software — charliermarsh · 2026-10-26
- Dev testing one skill across Opus 5.5 and Sol 6.1 finds instruction following 'all over the map' — lucasmeijer · 2026-10-26
- supermemory unveils learner-1, claiming continual learning for any model — julianweisser · 2026-10-26
- Anthropic says Opus 5.5 cuts cost per task by ~40% vs Opus 5 — taufiqintech · 2026-10-26
- AI search APIs ranked: top 12 within 7 points, yet prices differ nearly 7x — FinanceYF5 · 2026-10-26