Long agent tasks expose failures that benchmarks can miss
A model can score well on short tests but still struggle during agent work that lasts for hours. It may follow the first rules at the start, then forget them after about 40 minutes and undo earlier work without making the mistake obvious. Benchmark scores often do not warn about this because short, isolated tasks and long, steady work are different skills.
Long jobs such as a large code migration can reveal whether one early missed detail causes trouble much later. Many tested models lost track during this kind of work. GLM-5.2 kept its earlier choices more consistently and finished the long task without fighting its own prior decisions.
It is not presented as the strongest model for hard one-off questions, where large may still do better.
Key points
- Short benchmark scores may miss failures that appear in multi-hour .
- Long tasks can expose rule forgetting and accidental undoing of earlier work.
- Large code are a practical for agent .
- GLM-5.2 handled one long test better than expected for an .
- Large may still be stronger on difficult single questions.