Sub-Agent Max Reasoning Setting Drives Up Costs
TansuYegen · x · 2026-07-12
Someone mentioned that sub-agents in Sol default to max reasoning, prompting them to tweak settings to see if the cost experience improves. A previously heavy sub-agent that triggered too frequently has been fixed, but this highlights that it didn't actually need max reasoning, meaning it theoretically shouldn't have been so expensive.
A reply added some hands-on feedback: in a few large sessions, Fable's output quality was better, but the author still wouldn't use it because the cost/speed tradeoff isn't worth it. As long as the generated code lacks maintainability, they have to spend time managing the output manually, making the latency penalty unjustifiable.
Related event: GPT-5.6 Subagent Mechanism Sparks Cost and Over-triggering Controversy(5 posts)→
More from coding & agent
- One CLAUDE.md snippet stops Claude Code and Codex from using deprecated APIs — umeshonai · 2026-10-26
- HarnessMM paper cuts agent token use 87.9% on DeepSeek harness by reusing decisions — blaizedsouza · 2026-10-26
- NVIDIA open-sources SkillSpector, a security scanner for AI agent skills — NVIDIA · 2026-10-26
- Huashu Design 3.0 turns one agent prompt into deliverable design, 26k stars — alchaincyf · 2026-10-26
- OpenBot is a local-first desktop workspace for persistent AI teammates — nightly-labs · 2026-10-26
- Open-source agentic social scheduler Postiz hits 38k GitHub stars — gitroomhq · 2026-10-26