**Benchmark Fatigue**

> Benchmark scores for GPT-5.6, Fable 5.1, and Opus 5 don’t translate to real-world performance. Look under the hood, and you&\#039;ll find self-graded tests, missing numbers, and pure marketing spin. When new AI models hit the scene, companies often cite their scores on benchmarks such as BioMysteryBench, GeneBench, and Terminal-Bench as evidence of progress. But what do these tests evaluate, and do the scores even mat...

**来源信息**
- **来源**：Hacker News：AI 热帖
- **分类**：论文
- **发布时间**：2026-09-16 06:46（北京时间）
- **原文**：[打开原文](https://www.pcmag.com/opinions/why-ai-benchmarks-are-total-bs-and-how-openai-and-anthropic-use-them-to)