CodeRabbit measured GPT-6 Astra at 61.3% actionable bug coverage against Opus 5's 50.2%, at $1.50 per task versus Sol's $0.60

CodeRabbit·high signalXBlueskyLinkedInCopy link

CodeRabbit's 2026-09-04 evaluation scored models on actionable bug coverage against labeled bugs: Astra 61.3%, GPT-5.6 Sol 59.0%, Opus 5 50.2%, with the gap widening on cross-file reviews (Astra 57.1%, Sol 47.6%, Opus 5 42.9%). At a fixed 100K input / 10K output task the cost ladder is Astra $1.50, Sol $0.60, Terra $0.32, Luna $0.032, so Astra is 2.5x Sol and 47x Luna for a 4% overall coverage gain. They also flag the retention asymmetry: Astra supports zero data retention for eligible customers while Fable defaults to 30-day retention unless you are on Enterprise Frontier Safeguards.