Daily token badge undercounts usage: claude-fable-5 excluded from headline sum but counted in model breakdown %

Summary

The "N.NM today" token badge undercounts total usage because it appears to exclude the claude-fable-5 model from the headline sum, while the model breakdown percentage next to it DOES include Fable. The headline total and the percentage breakdown use inconsistent aggregation scopes.

Version / environment

• Claude Code 2.1.201
• macOS (local TUI), no custom statusLine configured (this is the native badge)

Expected

Headline "today" tokens = sum over ALL models used today (including claude-fable-5), consistent with the "Opus 73%" model breakdown shown right beside it.

Actual

Badge at 22:33 JST showed: 455.1M today · 内訳 Opus 73% · subagent 2.9%.

Aggregating my own local transcripts (~/.claude/projects/**/*.jsonl, cumulative at the same instant, summing input + output + cache_creation + cache_read):

| Scope | Tokens | Note |
| --- | --- | --- |
| Opus 4.8 only | 459.5M | matches the 455.1M headline |
| Fable 5 | 158.7M | NOT reflected in the headline |
| True grand total | 618.2M | actual usage today |

The breakdown "Opus 73%" matches Opus's share of the full total (74.3% of 618M) — i.e. the percentage denominator includes Fable, but the headline number does not.

Net effect:

• Headline underreports by ~26% (455M shown vs 618M actual).
• The headline stays visibly "stuck" while a concurrent Fable 5 session keeps consuming, because Fable tokens never move it.
• subagent 2.9% also disagrees with the measured subagent share (~0.8%), suggesting the headline-side aggregation scope is off more generally.

Repro / hypothesis

Run sessions on both an Opus model and claude-fable-5 in the same day, then watch the daily token badge. Likely cause: the headline usage aggregator's model/cost table doesn't yet include the newly added claude-fable-5 id, so its tokens are dropped from the sum, while the raw model-share breakdown still counts them.

Not a duplicate of

• #67506 (Fable 5 consumption vs its description) — that's about the magnitude Fable consumes; this is about Fable being omitted from the daily badge headline sum.
• #66756 (usage metrics inconsistent app vs web) — this is an inconsistency within the TUI badge itself (headline vs its own adjacent breakdown %).
• #73399 / #67171 / #73710 (context-window % using a 200K denominator for 1M/[1m] models) — those are about the context meter, not the daily token-usage badge.