# What are benchmarks where even the most recent frontier models barely make a dent or outperform previous ones only minimally? — retained record

> Retained reddit record from reddit:singularity.

Canonical: https://highsignal.app/data/records/998286e11c2a8298

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:singularity)  published 2026-11-10 17:40 ingested 2026-11-11 02:39

[original source ↗](https://www.reddit.com/r/singularity/comments/1xstb50/what_are_benchmarks_where_even_the_most_recent/)

## retained text



```
Many benchmarks now see scores in at least the mid-twenties (the ones most recently reported by Anthropic for Fable 5.2 are all above 40%) if we set aside ELO-scored ones,

What are challenging benchmarks with low top percentages and/or where the needle has been barely moving over the past months?

I'll start with [**SWE-sweep**](https://swesweep.com/) (mascot pictured), which deals with bug discovery and remediation without any hints, with a current top score of 9,5% (Astra). Fable 5.2 hasn't been tested yet, but Fable 5.1 scored 2.9%.
```



document fetched 2026-11-11 00:30 · canonical https://reddit.com/r/singularity/comments/1xstb50/what_are_benchmarks_where_even_the_most_recent
