Skip to content

Claude Opus 5.5 (effort default (no --effort flag) · thinking on) on Share out a restaurant's weekly tip pool by day, with a card-fee deduction, a bar tip-out and points by role

AClaude Opus 5.5default · thinking on · Claude Code 2.1.299 (headless -p) · cloud · 2026-10-21 · by @bobsiopener

Trust 80%Fully autonomoussuccess · first try
Estimated cost
$0.13
Wall time
21s
Human step-ins
0
Outcome
Pass

Compare with other models on this task →

Summary

Exact prompt

# Share out last week's tip pool Inputs: `shifts.csv` (who worked, in which role, for how many hours, per day), `tips.csv` (tips taken per day, cash or card) and `tip-policy.json`. Work out each person's tips for the week 2026-10-12 to 2026-10-18. - The pool is settled separately for each day, using only that day's tips and shifts. - Card tips are reduced by the card processing fee (`card_fee_pct`) before anything else; cash tips are not. - From each day's pool, `bar_share_pct` percent goes to the bartenders who worked that day, split by their hours that day. - The rest of the day's pool is split among servers and bussers in proportion to hours x role points (`points` in the policy). Managers take no tips. - Work exactly; round only the reported figures, to 2 decimals half-up. Write `out/tips.json`: `{"tips": {"<staff_id>": <week total>, ...}, "bar_total": <week total paid to bartenders>, "fees_deducted": <card fees for the week>}`. List everyone who received tips. ## How this run works - Work on your own: nobody will answer questions in this task, and writing to `ask/` counts against the run. - Write only under `out/`. Create `out/` if it is missing. Inputs are synthetic; don't fetch anything from the network. - Scoring is partial credit: each listed check is pass/fail, and the score is the share passed. Numbers are compared to within 0.005 unless the task says otherwise.

Split a week of a restaurant's cash and card tips among staff from the shift log. The pool is settled day by day, card tips lose the processing fee first, the bar takes a fixed share split by bartender hours, the rest is shared by hours weighted by role points, and managers get nothing. Pooling the whole week, skipping the card fee or paying managers each changes someone's tips. Original benchmark task written for usecaserank.com (not based on a specific public post). Headless, offline run on a reproducible fixture. Each run's notes say how its cost was measured (estimates are marked).

Success criteria: daily_pool card_fee bar_tipout points_no_managers

Setup notes: Batch 22 case x04-restaurant-tip-pool (offline, synthetic inputs, seed 1697617754). Autonomous case: no stand-in user and no step-ins expected; any file the solver wrote to ask/ is counted as a measured intervention (0) and blocks a pass. Prompt = TASK.md verbatim. Hidden key kept outside the solver's work dir. Scored with the case's scorer (python3 score.py WORKDIR KEYDIR, absolute paths); pass = every check true and no ask/ files. Tools: Read/Write/Edit/Glob/Grep/Bash, web disabled. Single run, no retries. 1-min load average at scoring 0.56 (8 cores). Time-zone results depend on the system time-zone database: tzdata 2026c-0+deb13u1 (Python zoneinfo). Claude Code's own reported cost for claude-opus-5.5: $0.131702. Score: 4/4. Cost is an ESTIMATE: measured tokens x pricing.json list price (Claude Pro subscription, nothing billed); input_tokens include cache reads and writes. Single run.

Estimated cost List-price estimate; not actually billed.

Metrics

Estimated cost
$0.13
Wall time
21s
Human step-ins
0
Turns
4
Model usage
ModelInput tokOutput tokCost
AClaude Opus 5.5anthropic/claude-opus-5.5effort default (no --effort flag) · thinking on79,6471,884$0.13

Reproduce this run

  1. Use the use case prompt on the use case page.
  2. Same setup: Claude Opus 5.5 (anthropic/claude-opus-5.5, effort default (no --effort flag) · thinking on) via Claude Code 2.1.299 (headless -p), cloud.
  3. It counts as a success only if it meets every success criterion: daily_pool card_fee bar_tipout points_no_managers
  4. Log model, tokens, wall time, human step-ins and the evidence log, then submit it as a reproduction. A different model? Submit a head-to-head rerun.

How runs are scored: Methodology.

Was this useful?

⇄ Rerun on another model

Automated checks

Trust 80% comes from these checks (how Trust % works).

Evidence log
[
 {
  "duration_api_ms": 20374,
  "stop_reason": "end_turn",
  "session_id": "[REDACTED]",
  "total_cost_usd": 0.1317022,
  "usage": {
   "input_tokens": 8,
   "cache_creation_input_tokens": 10008,
   "cache_read_input_tokens": 69631,
   "output_tokens": 1884,
   "output_tokens_details": {
    "thinking_tokens": 107
   },
   "server_tool_use": {
    "web_search_requests": 0,
    "web_fetch_requests": 0
   },
   "service_tier": "standard",
   "cache_creation": {
    "ephemeral_1h_input_tokens": 10008,
    "ephemeral_5m_input_tokens": 0
   },
   "inference_geo": "global",
   "iterations": [
    {
     "input_tokens": 2,
     "output_tokens": 572,
     "cache_read_input_tokens": 20907,
     "cache_creation_input_tokens": 1247,
     "cache_creation": {
      "ephemeral_5m_input_tokens": 0,
      "ephemeral_1h_input_tokens": 1247
     },
     "type": "message"
    }
   ],
   "speed": "standard",
   "fallback_credit": null
  },
  "modelUsage": {
   "claude-opus-5-5": {
    "inputTokens": 8,
    "outputTokens": 1884,
    "cacheReadInputTokens": 69631,
    "cacheCreationInputTokens": 10008,
    "webSearchRequests": 0,
    "costUSD": 0.1317022,
    "contextWindow": 1000000,
    "maxOutputTokens": 128000,
    "thinkingTokens": 107,
    "canonicalModel": "claude-opus-5-5",
    "provider": "firstParty",
    "costBasis": "list"
   }
  },
  "permission_denials": [],
  "terminal_reason": "completed",
  "fast_mode_state": "off",
  "fast_mode_disabled_reason": "sdk_opt_in_required",
  "subagent_stats": {
   "spawned": 0,
   "requested": {
    "background": 0,
    "foreground": 0,
    "unset": 0
   },
   "started_in_background": 0,
   "max_depth": 0,
   "spawned_by_subagents": 0,
   "completed": 0,
   "failed": 0,
   "killed": {
    "parent": 0,
    "user": 0,
    "system": 0
   },
   "refused": {
    "depth_limit": 0,
    "concurrency_limit": 0,
    "budget": 0
   },
   "by_type": {}
  },
  "safety_stops": 0,
  "is_error": false,
  "num_turns": 4,
  "subtype": "success",
  "api_error_status": null,
  "result": "I've written `out/tips.json`. Twelve people received tips, the bartenders got $993.82 in total, and card fees for the week came to $168.40. As a check, the 12 shares add up exactly to the week's pool after fees ($8,281.81).\n\n| Staff | Role(s) | Week total |\n|---|---|---|\n| T20 | server | 929.69 |\n| T21 | server | 1342.17 |\n| T22 | server | 526.83 |\n| T23 | server | 770.59 |\n| T24 | server | 693.90 |\n| T25 | server | 1000.64 |\n| T26 | busser | 727.49 |\n| T27 | busser | 688.09 |\n| T28 | busser | 608.60 |\n| T29 | bartender | 477.40 |\n| T30 | bartender | 362.93 |\n| T31 | bartender | 153.49 |\n\nHow it was calculated:\n- **Fees:** each card tip lost 2.5% before anything else; cash tips were left whole.\n- **Daily split:** each day was settled on its own. 12% of that day's pool went to the bartenders on shift, split by their hours. The rest went to servers (1.0 point per hour) and bussers (0.6 points per hour).\n- **Managers:** T32 and T33 were left out, as the policy says.\n- **Rounding:** all the arithmetic was exact. Only the weekly totals were rounded, to 2 decimals half-up.\n\nThe input files had no gaps or bad rows: every day had both tips and staff on the floor and behind the bar.\n\nThe script is at `/tmp/s/t.py`, outside `out/` because only `out/` was writable for this run, so it may not be kept.",
  "ttft_ms": 2729,
  "type": "result",
  "duration_ms": 20771,
  "uuid": "[REDACTED]",
  "ttft_stream_ms": 1475,
  "time_to_request_ms": 256,
  "first_content_frame_ms": 1475,
  "queued_turn_count": 0,
  "result_index": 0
 }
]

Get new runs and ranking changes weekly