fix(usage): price Claude fast-mode requests at the fast rate

[t3dotgg]

Usage estimated Claude fast-mode requests at the standard rate. Fast mode bills at a multiple of standard: 2x for Opus 5.5 ($8/$40 per MTok), Opus 5, and Opus 4.8. So the page showed half the real cost for those requests.

Claude Code writes usage.speed: "fast" on these requests, and LiteLLM publishes the multiple in provider_specific_entry.fast. We read neither.

Fix

• parseClaudeLine sets fast on the record from usage.speed.
• parseRateTable reads provider_specific_entry.fast into fastMultiplier (default 1).
• priceUsage and cacheSavingsUsd take the record and apply the multiplier. Custom prices apply as entered.
• Scan cache v3 → v4, so cached transcripts re-parse once and pick up the flag.

No contract or UI change.

Scope

I compared our pricing against current ccusage and the live LiteLLM table. Base rates for Fable 5.1 (cache reads $0.25) and Opus 5.5 ($4/$20, cache reads $0.20) are already correct because we fetch LiteLLM live. Fast mode is the only gap that no open PR covers. These are left to open PRs:

• 1h cache writes, long-context tiers, and keeping the fullest snapshot per message: fix(usage): correct Claude cost accounting #9019
• Advisor-tool tokens: fix(server): include Claude advisor calls in usage totals #10488

Proof: On the real LiteLLM table, a fast Opus 5.5 line prices at $0.28 on main and $0.56 on this branch, which matches $8/$40. vp test run apps/server/src/usage/ passes (87 tests). Server tsc --noEmit, lint, and fmt pass.

Made by Claude Opus 5.5 (1M context) in Claude Code, running in T3 Code.

🤖 Generated with Claude Code

Closes #6306

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This production usage-metering change alters the calculated cost and cache-savings values for existing Claude fast-mode requests and changes the persisted scan-cache format. The logic is localized and tested, but billing-sensitive behavior requires human review.

^{You can add or adjust custom eligibility rules. Learn more.}

[coderabbitai]

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

✨ Finishing Touches

📝 Generate docstrings

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

---

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

ℹ️ The exact PR base did not have a successful artifact. Baseline uses the latest successful main measurement shown below.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.5 KiB	13.6 KiB	+50 B (+0.4%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.0 KiB	7.1 KiB	+16 B (+0.2%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.5 KiB	6.5 KiB	+34 B (+0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	56.3 KiB	56.3 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	10	10	0 (0.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	−12 B (−0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.1 KiB	7.1 KiB	+1 B (+0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.5 KiB	6.4 KiB	−13 B (−0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.0 KiB	57.0 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	9	9	0 (0.0%)	21	✅

Baseline: 408ff8a · PR result: d4ed85c · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 114.0 KiB
• Claude decoded thread snapshot: 114.7 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.