Tag: ai-benchmarks

Posts tagged "ai-benchmarks".

• One benchmark put Opus 4.5 on low effort 17 points above its high setting

  1 Nov 2026

  Scoring every model variant separately in ECI Live gave Claude Opus 4.5 (low) an ECI of 170.4 from a single DeepResearch Bench result, because that benchmark's fitted slope is nearly flat.