Skip to content

Claude Haiku 5.5 (effort default (no --effort flag) · thinking on) on Invoice a caterer's events with a large-party discount, weekday minimum spends, a service charge and VAT

AClaude Haiku 5.5default · thinking on · Claude Code 2.1.299 (headless -p) · cloud · 2026-10-22 · by @bobsiopener

Trust 80%Fully autonomoussuccess · first try
Estimated cost
$0.0034
Wall time
12s
Human step-ins
0
Outcome
Pass

Compare with other models on this task →

Summary

Exact prompt

# Invoice a caterer's events for November Inputs: `events.csv` (each booked event: date, guests, menu), `menus.json` (price per head for each menu) and `terms.json` (the weekday minimum spends). Every amount is in pounds. Invoice rules, per event: - Food = per-head price x guests, except that guests beyond the first 100 are charged at 10% off the per-head price. - If food is below the minimum spend for the event's weekday in terms.json, add a top-up so that food plus top-up equals the minimum; otherwise the top-up is 0. - Service charge is 12.5% of food plus any top-up. - VAT at 20% applies to food, top-up and service charge. - Round food, service charge and VAT each to the cent half-up as you compute them (later steps use the rounded figures); total = food + top-up + service charge + VAT. Write `out/invoices.json`: `{"food": {"<event_id>": <food>, ...}, "top_up": {...}, "service": {...}, "vat": {...}, "total": {...}, "grand_total": <sum of the totals>}` with every event in `events.csv`. ## 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.

Invoice a month of catered events from guest counts and menu prices. Guests beyond the first 100 get 10% off, events below their weekday's minimum spend get a top-up, a 12.5% service charge covers food and top-up, and VAT covers everything. Discounting every guest, leaving the top-up out of the service charge or leaving service out of VAT each change invoices. 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: tier_discount minimum_spend service_charge vat_and_totals

Setup notes: Batch 24 case z01-event-catering-invoices (offline, synthetic inputs, seed 269412003). 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 0.53 (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-haiku-5.5: $0.003414. 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
$0.0034
Wall time
12s
Human step-ins
0
Turns
3
Model usage
ModelInput tokOutput tokCost
AClaude Haiku 5.5anthropic/claude-haiku-5.5effort default (no --effort flag) · thinking on60,8012,162$0.0034

Reproduce this run

  1. Use the use case prompt on the use case page.
  2. Same setup: Claude Haiku 5.5 (anthropic/claude-haiku-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: tier_discount minimum_spend service_charge vat_and_totals
  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": 11216,
  "stop_reason": "end_turn",
  "session_id": "[REDACTED]",
  "total_cost_usd": 0.00341437,
  "usage": {
   "input_tokens": 6,
   "cache_creation_input_tokens": 9078,
   "cache_read_input_tokens": 51717,
   "output_tokens": 2162,
   "output_tokens_details": {
    "thinking_tokens": 664
   },
   "server_tool_use": {
    "web_search_requests": 0,
    "web_fetch_requests": 0
   },
   "service_tier": "standard",
   "cache_creation": {
    "ephemeral_1h_input_tokens": 9078,
    "ephemeral_5m_input_tokens": 0
   },
   "inference_geo": "global",
   "iterations": [
    {
     "input_tokens": 2,
     "output_tokens": 889,
     "cache_read_input_tokens": 20036,
     "cache_creation_input_tokens": 1581,
     "cache_creation": {
      "ephemeral_5m_input_tokens": 0,
      "ephemeral_1h_input_tokens": 1581
     },
     "type": "message"
    }
   ],
   "speed": "standard",
   "fallback_credit": null
  },
  "modelUsage": {
   "claude-haiku-5-5": {
    "inputTokens": 6,
    "outputTokens": 2162,
    "cacheReadInputTokens": 51717,
    "cacheCreationInputTokens": 9078,
    "webSearchRequests": 0,
    "costUSD": 0.00341437,
    "contextWindow": 1000000,
    "maxOutputTokens": 128000,
    "thinkingTokens": 664,
    "canonicalModel": "claude-haiku-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": 3,
  "subtype": "success",
  "api_error_status": null,
  "result": "I wrote `out/invoices.json` with all eight events (EV-3105 to EV-3149). The grand total is \u00a340,700.15.\n\nThe per-event figures:\n\n| Event | Weekday | Food | Top-up | Service | VAT | Total |\n|---|---|---|---|---|---|---|\n| EV-3105 | Thu | 4794.40 | 0.00 | 599.30 | 1078.74 | 6472.44 |\n| EV-3113 | Thu | 4993.81 | 0.00 | 624.23 | 1123.61 | 6741.65 |\n| EV-3120 | Sat | 2397.20 | 102.80 | 312.50 | 562.50 | 3375.00 |\n| EV-3124 | Tue | 3930.22 | 0.00 | 491.28 | 884.30 | 5305.80 |\n| EV-3131 | Thu | 2343.25 | 0.00 | 292.91 | 527.23 | 3163.39 |\n| EV-3139 | Tue | 6370.56 | 0.00 | 796.32 | 1433.38 | 8600.26 |\n| EV-3143 | Fri | 702.65 | 1097.35 | 225.00 | 405.00 | 2430.00 |\n| EV-3149 | Thu | 3416.01 | 0.00 | 427.00 | 768.60 | 4611.61 |\n\nChoices I made:\n- **Weekday** comes from each event's date in 2026, so the minimum spend is the one for that day of the week.\n- **Discount** applies only to guests above 100. EV-3113 (186 guests) is the only event affected.\n- **VAT** is 20% of food plus top-up plus service charge, rounded to the cent.\n- **Totals** are food + top-up + service + VAT, and the grand total sums those per-event totals.\n- **Output values** are written as JSON numbers, which carry the cent values exactly.\n\nI did not check the contents of `TASK.md` against the brief above, and I did not run an independent check of the arithmetic beyond recomputing one event by hand.",
  "ttft_ms": 1570,
  "type": "result",
  "duration_ms": 11554,
  "uuid": "[REDACTED]",
  "ttft_stream_ms": 1322,
  "time_to_request_ms": 244,
  "first_content_frame_ms": 1322,
  "queued_turn_count": 0,
  "result_index": 0
 }
]

Get new runs and ranking changes weekly