Low Reasoning Effort Slashes AI Costs

🔥 What happened A new Vals Index study found that at their lowest reasoning setting, four of seven frontier models kept at least 80% of their score for half the cost or less. Claude Opus 5.5 held 80% of its score at just 16% of the cost. 💡 Why it matters In 22 of 56 model-benchmark pairs, a cheaper setting matched the leaderboard score. Even perfect per-task tuning adds only 6 points — at 64% of the cost. Running the lowest setting everywhere costs just 29%. ⚡ Our take Defaulting to max reasoning is burning money for nothing. If you're not throttling per task, you're subsidizing your bill, not your quality.