AnalysisAI ModelsSeptember 26, 2026

Reddit user reruns Terminal-Bench 2.1 on Luna 5.6 and Luna 6

Read original source →reddit.com

A Reddit user ran 100 Terminal-Bench 2.1 slots comparing GPT-5.6 Luna and GPT-6 Luna, then repeated the comparison two days later to check whether the large gap was a bad run. They report the Luna 6 results still look like a joke.

1 source

More stories today

Open the live feed