廉价档四强对比：M3.1 Flash AA 42 分、$0.15/$0.60，与 DeepSeek、Haiku 5.5
作者按 AA 指数与每百万 token 价格排开 DeepSeek V4.1 Flash（45 分，$0.30/$1.20）、Haiku 5.5（43 分，$0.10/$0.50）、MiniMax M3.1 Flash（42 分，$0.15/$0.60）和 Mercury 3，主张用自家仓库测试挑最便宜的可用模型。
the cheap tier just became a f*cking four-way fight artificial analysis index · price per 1M in / out: → deepseek v4.1 flash 1017: 45 · $0.30 / $1.20 (half off-peak) → haiku 5.5: 43 · $0.10 / $0.50 (under a 100K prompt) → minimax m3.1 flash: 42 · $0.15 / $0.60 → mercury 3: new this week, inception says it beats haiku 5.5 at 1,000+ tok/s 3 index points apart, 3x apart on input price for grep, summaries and routing, the cheapest model that passes your own tests wins the prompt in the picture runs that test on your repo the scout setup is in the article below which one is your scout model right now?
这条是引用着另一条写的：x.com/lostinazeroy/status/2108531017979920732