GoKawiilBenchmark test finds GPT-5.6 Luna catches fewer bugs than GPT-6 Astra but at 28x lower cost
A new benchmark comparing OpenAI's GPT-5.6 Luna and GPT-6 Astra on code review found Luna verified 69 bugs across 50 pull requests versus Astra's 92, while costing roughly 28 times less per review. Luna also produced more false positives, with 24 of 93 flagged issues failing verification compared to Astra's 4 of 96, and it caught fewer security-related bugs.
GoKawiil's interpretation of the reporting above, not reported fact.
As AI code review tools proliferate, teams face a real tradeoff between cost and accuracy rather than a clear winner. This test suggests cheaper models like Luna can handle routine correctness checks affordably, but relying on them alone for security-sensitive code like authentication or permissions carries measurable risk.
- Luna costs about 28x less per pull request review than Astra but misses more bugs
- Luna's findings had a higher false-positive rate, with roughly 26% failing verification versus 4% for Astra
- Astra caught nearly twice as many verified security bugs as Luna, making it riskier to rely on cheaper models for sensitive code
Source: entelligence.ai, 2026-09-14
Published there as: “GPT-5.6 Luna vs. GPT-6 Astra: Is a $1.20 Model Good Enough for Code Review?”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.