You've hit your session limit · resets 4pm (America/Los_Angeles)
Bug Description
Model tiering failure burned my Fable 5 usage limit. I explicitly instructed
the orchestrator in my first task message to delegate subtasks to cheaper/faster
models. It ignored this and launched a multi-agent research workflow (~60 subagents)
where ALL subagents silently inherited the session model (Fable 5), because the
Workflow/Agent tools default to inheriting the orchestrator's model unless
overridden per-agent. ~15-20 Fable subagent calls burned on mechanical work
(web search, page fetching) before I caught it and forced a stop.
Suggestions: (1) when the session model is a limited premium model, subagent
fan-out should warn or default to a cheaper tier; (2) surface per-model token
spend during workflow runs so users see the burn early.
Date: 2026-07-04, session on Claude Code with Fable 5.
Environment Info
- Platform: darwin
- Terminal: ghostty
- Version: 2.1.201
- Feedback ID: 7f151af2-8db2-4a90-8cb9-3ab8b6a663b8
Errors
[]This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗