Opus 5.5 leads at $5.98/task while GPT-6 Luna costs just $0.07 — a 85x gap

haider1 · x · 2026-09-23

Benchmarks comparing newest models on cost per task: Opus 5.5 leads with 58 points at $5.98/task, GPT-6 sol hits 48 at $1.06, and GPT-6 luna scores 37 at just $0.07. Opus also consumes over 2x the output tokens of GPT-6 sol. The author questions the AA benchmark's credibility, noting GPT-6 sol ranks on par with Muse Spark 1.3 and only 3 points behind Opus 5.

Original post →

More from Models

Models channel →