Problem
The ucode usage command groups token usage by coding agent (Claude Code, Codex, etc.) and merges all model usage into a single total per agent. The per-model breakdown in the Top models this week summary is misleading — it appears to show the agent's total tokens next to each model, rather than the tokens used by that specific model.
Example
From a recent run:
Top models this week: Claude Haiku 4.5 (237.0K), Claude Opus 4.7 (237.0K), GPT-5 (13.3K)
Claude Code had 237.0K tokens in total across both models. But the summary shows Claude Haiku 4.5 (237.0K) AND Claude Opus 4.7 (237.0K), which implies each model used 237K independently (double-counting).
The actual breakdown should be something like:
Top models this week: Claude Opus 4.7 (236K), Claude Haiku 4.5 (920), GPT-5 (13.3K)
Where it shows up
The per-agent daily table (Claude Code · Last 7 Days) also lumps models together in the Models column (e.g. Claude Haiku 4.5, Claude Opus 4.7) without per-model token counts, so there's no way to see the actual split for a given day.
Requested change
- Fix the
Top models this week summary so each model shows only the tokens consumed by that model (not the agent total).
- Consider breaking out per-model token counts in the daily table below — either as additional columns, or by showing tokens next to each model name.
Problem
The
ucode usagecommand groups token usage by coding agent (Claude Code, Codex, etc.) and merges all model usage into a single total per agent. The per-model breakdown in theTop models this weeksummary is misleading — it appears to show the agent's total tokens next to each model, rather than the tokens used by that specific model.Example
From a recent run:
Top models this week: Claude Haiku 4.5 (237.0K), Claude Opus 4.7 (237.0K), GPT-5 (13.3K)
Claude Code had 237.0K tokens in total across both models. But the summary shows
Claude Haiku 4.5 (237.0K)ANDClaude Opus 4.7 (237.0K), which implies each model used 237K independently (double-counting).The actual breakdown should be something like:
Top models this week: Claude Opus 4.7 (236K), Claude Haiku 4.5 (920), GPT-5 (13.3K)
Where it shows up
The per-agent daily table (
Claude Code · Last 7 Days) also lumps models together in theModelscolumn (e.g.Claude Haiku 4.5, Claude Opus 4.7) without per-model token counts, so there's no way to see the actual split for a given day.Requested change
Top models this weeksummary so each model shows only the tokens consumed by that model (not the agent total).