Airbnb公布生产级分类基准：微调小模型以1%成本媲美前沿模型
Start here
2026年11月6日，Airbnb公布了生产级大型语言模型（LLM）分类器的基准评测结果。在真实生产分类任务中，Airbnb对比了微调后的 E2B 模型与前沿商业模型 API。测试结果显示，经过微调的小模型在 100 毫秒内达到了前沿商业模型的分类质量，而运行成本仅约为后者的百分之一。
AI-generated from reports · Updated 3 hours ago
Report timeline
Follow the reports to see different sides of the story.
- Hacker News · AISelectedAirbnb production classification benchmark: Fine-tuned small models match frontier models at 1% of the cost
In a real-world production classification task, Airbnb benchmarked fine-tuned E2B models against frontier commercial model APIs. The results showed that the fine-tuned model matched the quality of frontier models within 100ms, at roughly 1% of the cost.
Story heat trend
Current heat 9·Comparable peak 10（Nov 6 22:00）·Comparable change, last 24 hours–
Trends compare only the same parties observed continuously, so the scope can be narrower than current heat counts. Move the pointer or tap the chart to see hourly heat; use the left and right arrow keys on a keyboard.