$1.2模型能替代$50做代码审查？
推荐指数 67.0 NO. 012 · 2026.09.15
发布2026/09/14Score57Comments71
为什么值得看
GPT-5.6 Luna以28倍成本优势完成代码审查，50个PR中发现69个bug，但安全漏洞检出率仅为GPT-6 Astra的一半。对预算敏感的团队，这意味着日常代码审查可以大幅降本，但安全关键场景仍需高配模型兜底。
编辑判断
这个测试揭示了一个被忽视的成本结构问题：大多数团队的代码审查预算其实花在了过度检测上。Luna 24%的误报率 vs Astra 4%的差距，在工程实践中往往可以通过简单的规则过滤或人工抽检消化掉，而28倍的成本差却是实打实的现金流。
更关键的信号是安全漏洞的检出差距（9 vs 19）。这暗示了一个分层架构的可行性——用 Luna 做日常扫描，仅在涉及安全敏感文件或外部依赖变更时触发 Astra。GitHub Actions 的条件路由已经支持这种逻辑，值得团队本周就验证。
对 AI infra 创业者来说，这里的机会不是做更好的模型，而是做智能路由层：根据代码变更的语义特征动态选择模型，把平均成本压到 Luna 水平线同时保持 Astra 的安全底线。
社区反馈
意见分歧 69 条评论
核心争论：廉价模型能否替代高端模型做代码审查，还是应追求质量而非极致降本
We're using an even cheaper model (K2.7 right now) and our PRs cost more to review (still < $1). I imagine we have a lot more going on in terms of review, even after moving the context gathering and comment posting responsibilities to scripts that run before/after the agent. It's more of a team
Per the article the luna review cost $0.004 and astra cost $0.113. The headline is per million tokens
I understand that, but sub-penny costs means you are hardly doing anything, even at the per-million token prices (which we pay lower values for, but more per-review overall). I cannot imagine they are doing as good of a review as they could be doing, which is to say minimizing pr review spend is not