Claude Code's /usage Stats tab overstates tokens ~2x; 5 GitHub reports since Aug, still unfixed
Background-Basis-672 · reddit · 2026-09-16
A user verified that Claude Code's /usage Stats tab shows roughly double the real token usage: it writes one transcript row per content block, and each row repeats the entire call's usage. The adjacent Usage tab and claude -p --output-format dedupe correctly (1,647 requests vs 4,430 rows in the original bug report), and Anthropic's own docs say to dedupe on message id.
The issue has been reported on GitHub five times since August 2025 (#6805, #87303, #90991, #91775, #94224) with no human reply. Billing and limits are server-side, so nobody is overcharged — but the number skews how users judge plan value and compare across vendors. The author provides a read-only counting script and a full write-up.
More from coding & agent
- Microsoft's MemSteer turns unlabeled user histories into agent memory supervision — MicrosoftResearch · 2026-11-02
- She fired herself from her own Instagram: AI agent runs the account, views up 82.3% — heyshrutimishra · 2026-11-02
- Sourcegraph: AI writes code faster than teams can review it, driving more outages — AI Engineer · 2026-11-02
- Tsinghua & Microsoft's HarnessMM cuts agent token use by up to 87.93% while raising success rate — rohanpaul_ai · 2026-11-02
- GitHub launches Agent HQ: Anthropic, OpenAI, Google coding agents in Copilot — emmanuelvivier · 2026-11-02
- DeepSeek open-sources Super-Agent, a coding agent that reads team chats to minimize human input — emmanuelvivier · 2026-11-02