"requires usage credits" error for Fable 5 despite all quotas well under limit (session 3%, week 26%, week Fable 24%)
Title: "requires usage credits" error for Fable 5 despite all quotas well under limit (session 3%, week 26%, week Fable 24%)
Body:
Bug description
While using Fable 5 in a Claude Code CLI session, I got a dialog claiming I need to purchase/pay for usage credits. This contradicts the /status output from the very same session, which shows plenty of headroom left on every relevant quota:
Current session
█▌ 3% used
Resets 4:50pm (Europe/Berlin)
Current week (all models)
█████████████ 26% used
Resets Jul 21, 6pm (Europe/Berlin)
Current week (Fable)
████████████ 24% used
Resets Jul 21, 6pm (Europe/Berlin)
Session cost/usage summary for context:
Total cost: $53.30
Total duration (API): 1h 56m 10s
Usage by model:
claude-haiku-4-5: 1.3k input, 51 output, 0 cache read, 0 cache write ($0.0016)
claude-opus-4-8: 50.1k input, 299.6k output, 23.7m cache read, 1.8m cache write ($36.55)
claude-sonnet-5: 7.0k input, 111.7k output, 10.2m cache read, 647.6k cache write ($8.13)
claude-fable-5: 1.1k input, 36.2k output, 2.0m cache read, 238.5k cache write ($8.61)
Steps to reproduce
- In an active CLI session, switch model to Fable 5 via
/model. - Continue the conversation — a dialog appears stating usage credits are required / must be purchased.
- Switch back to Sonnet 5 via
/model(confirmed via "Kept model as Sonnet 5") — the same credits-required condition persists in the session, even though/modelreports Sonnet is active. - Run
/statusin the same session — all three quota bars (session, weekly all-models, weekly Fable) are well below 100%, as shown above.
Expected behavior
No "usage credits required" prompt should appear while every quota shown in /status has headroom remaining. If Fable 5 draws from a separate, unlisted credit pool distinct from the weekly Fable bar shown in /status, that pool and its remaining balance should be surfaced somewhere — currently there is no way to see why it thinks credits are needed.
Environment
- Claude Code CLI
- Session includes model switches: Fable 5 → Sonnet 5 (confirmed via
/model) - Web interface (claude.ai) usage view for Fable shows 24% used, consistent with the CLI's own
/status, further confirming the credits-required dialog is inconsistent with all visible usage data.
Impact
Blocks continued use of the session/model despite no quota being exhausted; inconsistent messaging between the credits dialog, /status, and the web usage view is confusing and makes it unclear whether this is a real billing block or a display/logic bug.
This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗