Claude Benchmarks Revealed as Mixed Output with Opus 4.8

soumitrashukla9 · x · 2026-07-04

danshipper points out that the debated model is actually identical to the original version and not a new release. However, its fallback rate to Opus 4.8 has slightly increased, meaning recent benchmark results actually measure a hybrid output of two models, potentially skewing assessments of its true performance.

Original post →

More from Models

Models channel →