fix(code-review): drop model:sonnet override — 1M-context billing footgun (v2.37.1) - #169
Merged
Merged
Conversation
…tgun PR #167 added `model: sonnet` to the /start, /deep, /shallow command frontmatter to run the orchestrator on Sonnet. But a per-command model override inherits the session's context-window tier: on a 1M-context session (e.g. a default of "Opus 4.8 (1M context)") it resolves to Sonnet-with-1M, which bills as extra pay-as-you-go API usage OUTSIDE a Pro/Max subscription (Sonnet 4.6 has no long-context rate premium, but its 1M tier is not subscription-included the way Opus's is). There is no per-command way to opt out of the session's 1M tier (known Claude Code limitation: anthropics/claude-code#45847, #45169), so the override inverted the intended saving for anyone on a 1M-default session. Remove the override; the orchestrator runs on the session model again. The model-independent turn-and-context discipline rules from v2.37.0 are retained. start.md and README.md now document running /code-review from a standard-context Sonnet session (`/model sonnet`) as the supported way to get the cheaper orchestrator, and explain why the frontmatter override is deliberately avoided so it is not re-added. Version 2.37.1 (PATCH). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The bug
PR #167 added
model: sonnetto the/start,/deep,/shallowcommand frontmatter to run the orchestrator on Sonnet. But a per-command model override inherits the session's context-window tier — so on a 1M-context session (e.g. a default of "Opus 4.8 (1M context)") the override resolves toclaude-sonnet-4-6[1m], which bills as extra pay-as-you-go API usage outside a Pro/Max subscription.Verified against current Claude Code behavior + docs:
So for anyone whose default session is a 1M model, #167 silently moved the orchestrator out of the subscription into billable API usage — and Opus-1M (included) is actually cheaper for them than Sonnet-1M (extra). The "optimization" inverted.
The fix
Remove the
model: sonnetoverride from all three commands. The orchestrator runs on the session model again.start.mdandREADME.mdnow document the supported path to the cheaper orchestrator: run/code-reviewfrom a standard-context Sonnet session (/model sonnet), where Sonnet bills normally and is genuinely cheaper than Opus. The docs also explain why the frontmatter override is deliberately avoided, so it isn't re-added.Scope / sequencing
6-file change (frontmatter + docs + version + CHANGELOG), no code. 2.37.0 → 2.37.1 (PATCH).
Branched from
main(which has #167). Independent of the open #168 (premise category excision, v3.0.0) — they touch different sections ofstart.md. This hotfix should merge first (it's an active billing regression); #168 then rebases and re-bumps from 2.37.1.🤖 Generated with Claude Code