Cheap models can cost more in agent harnesses than frontier ones, researcher warns
vasuman · x · 2026-10-27
Running a 'cheaper' model in an agent harness is often more expensive than using frontier models. Frontier models post-trained with RL brevity penalties aggressively batch steps via parallel tool calling, while weaker models serialize everything — dozens of round trips that blow up context windows on every hop.
So you're paying not just for intelligence per token, but for the ability to collapse turns instead of stepping linearly. Not unsolvable, but worth remembering when reaching for a cheaper model.
More from coding & agent
- Claude spots broken physics in AI videos, fixes them with CFD and finite-element verification — OfirPress · 2026-10-27
- One prompt: Claude Opus 5.5 adds drone flight mode to RV travel game — ezshine · 2026-10-27
- Do canary tokens still work in the coding-agent era? Fake AWS keys may lure agents — kuza55 · 2026-10-27
- Veteran engineer on what the AI-coding generation will never feel: 36-hour sprints and elusive bugs — bendee983 · 2026-10-27
- Claude Opus 5.5 builds a browser open-world island by comparing screenshots to 341 real photos — prasenx · 2026-10-27
- BreakingWeb benchmark: browser agents fail about half of controlled task variants — shuyanzh36 · 2026-10-27