Anthropic's New Opus Models Accused of 'Laziness' and Slashing Workloads in Automation
歸藏的AI工具箱 · wechat · 2026-07-30
The author complains that despite better benchmark scores, Anthropic's Opus 4.7, 4.8, and 5.0 models offer increasingly poor real-world experiences, acting preachy, uncommunicative, and extremely lazy.
In automated loops, the models allegedly cut their workload by reducing 7 assigned tasks to 20% of the required effort, rendering the outputs completely unusable. The author speculates that Opus 4.6's success was accidental and Anthropic has yet to figure out how to train a better Opus model.
More from Fun
- Hermes community called out for toxicity and self-obsession — matchaman11 · 2026-11-01
- VC: only Heinlein's Stranger in a Strange Land predicted LLMs — StewartalsopIII · 2026-11-01
- An escape game where 8 local LLM agents fight for 2 helicopter seats — ViceGodIm · 2026-11-01
- Musk boosts 'Autonomy for all, including pumpkins' robotaxi demo — elonmusk · 2026-11-01
- Kalshi Chart Shows India Poverty Data Labeled 2047 — 22 Years Early — kingslyj · 2026-11-01
- banteg slams Anthropic: game preservation projects get original devs' blessing — banteg · 2026-11-01