0.8B 小模型,把 GPT-5.6 Sol xhigh 打了
🚨0.8B 小模型,把 GPT-5.6 Sol xhigh 打了。
Shopify CEO Tobi Lütke 今天晒了一个挺夸张的内部案例:他们在 Buyer Profile 这个高度垂直的任务上,把 Qwen3.5-0.8B 微调后,内部 Judge 得分做到 84.6,超过 GPT-5.6 Sol xhigh 的 83.0,也远高于此前线上 2B 模型的 77.2。
更离谱的是效率。
tobi lutke @tobi Training tiny models for special purpose use cases works so incredibly well if you have a great self improving recursive flywheel. Shopify ML team is on fire.
finetuned 0.8b model beats GPT 5.6-sol xhigh in this very specialized task.