Skip to content

Claude Haiku 5.5 (effort default (no --effort flag) · thinking on) on Bill a window cleaner's monthly round with add-on jobs, no-access call-outs, a direct debit discount and a minimum charge

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
$0.0039
Wall time
14s
Human step-ins
0
Outcome
Pass

Compare with other models on this task →

Summary

Exact prompt

# Bill a window cleaner's round for March Inputs: `visits.csv` (each visit: customer, date, house type, extras separated by `;`, outcome), `customers.json` (each customer's payment method) and `prices.json` (house-type prices, extra prices and the monthly minimum). Every amount is in pounds. Billing rules: - A cleaned visit (outcome cleaned) costs the house-type price plus the price of each listed extra, less 2.00 if the customer pays by direct_debit. - A no-access visit (outcome no_access) is charged a call-out of 50% of the house-type price (rounded to the cent half-up), with no extras and no direct debit discount. - If a customer's visits for the month add up to less than the monthly_minimum, add a top-up so the month comes to the minimum; the minimum applies per customer per month, not per visit. Write `out/round.json`: `{"visit_price": {"<visit_id>": <price>, ...}, "top_up": {"<customer>": <top-up, 0 if none>, ...}, "due": {"<customer>": <visits + top-up>, ...}, "grand_total": <sum of due>}` with every visit and every customer in `visits.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.

Bill a month of window cleaning visits. Cleaned visits cost the house-type price plus any extras, less a direct debit discount, while no-access visits cost half the house price with no extras or discount, and each customer pays at least a monthly minimum. Charging extras on no-access visits, discounting them or applying the minimum per visit each change the bills. 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: visit_prices monthly_minimum totals

Setup notes: Batch 25 case a05-window-cleaning-round (offline, synthetic inputs, seed 239628767). 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 0.27 (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: $0.00387. Score: 3/3. 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
$0.0039
Wall time
14s
Human step-ins
0
Turns
3
Model usage
ModelInput tokOutput tokCost
AClaude Haiku 5.5anthropic/claude-haiku-5.5effort default (no --effort flag) · thinking on62,6492,494$0.0039

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: visit_prices monthly_minimum 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": 12714,
  "stop_reason": "end_turn",
  "session_id": "[REDACTED]",
  "total_cost_usd": 0.00386979,
  "usage": {
   "input_tokens": 6,
   "cache_creation_input_tokens": 10504,
   "cache_read_input_tokens": 52139,
   "output_tokens": 2494,
   "output_tokens_details": {
    "thinking_tokens": 911
   },
   "server_tool_use": {
    "web_search_requests": 0,
    "web_fetch_requests": 0
   },
   "service_tier": "standard",
   "cache_creation": {
    "ephemeral_1h_input_tokens": 10504,
    "ephemeral_5m_input_tokens": 0
   },
   "inference_geo": "global",
   "iterations": [
    {
     "input_tokens": 2,
     "output_tokens": 997,
     "cache_read_input_tokens": 20434,
     "cache_creation_input_tokens": 2609,
     "cache_creation": {
      "ephemeral_5m_input_tokens": 0,
      "ephemeral_1h_input_tokens": 2609
     },
     "type": "message"
    }
   ],
   "speed": "standard",
   "fallback_credit": null
  },
  "modelUsage": {
   "claude-haiku-5-5": {
    "inputTokens": 6,
    "outputTokens": 2494,
    "cacheReadInputTokens": 52139,
    "cacheCreationInputTokens": 10504,
    "webSearchRequests": 0,
    "costUSD": 0.00386979,
    "contextWindow": 1000000,
    "maxOutputTokens": 128000,
    "thinkingTokens": 911,
    "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": "Wrote `out/round.json`. Grand total is \u00a3278.44.\n\n| Customer | Visits | Top-up | Due |\n|---|---|---|---|\n| 12 Elm Road | 52.70 | 0.00 | 52.70 |\n| Rose Cottage | 103.40 | 0.00 | 103.40 |\n| 7 Orchard Close | 9.34 | 5.66 | 15.00 |\n| Flat 2, Quay House | 35.76 | 0.00 | 35.76 |\n| 48 Station Rd | 56.58 | 0.00 | 56.58 |\n| The Old Forge | 8.94 | 6.06 | 15.00 |\n\nI computed the figures with a script using exact decimal arithmetic, then checked the customer totals by hand against the rules. Each visit is priced as follows:\n- **Cleaned visits:** house price plus extras, less \u00a32.00 for direct debit customers.\n- **No-access visits:** 50% of the house price, rounded half-up to the cent (all come to \u00a38.94 here). No extras or discount.\n- **Monthly minimum:** \u00a315.00 per customer, topped up where the visit total falls short.\n\nI wrote the output with `float` values, so `52.7` appears where `52.70` would be written in the spec. The numbers are equal within the 0.005 tolerance. I wrote nothing under `ask/` and nothing outside `out/`.",
  "ttft_ms": 1733,
  "type": "result",
  "duration_ms": 13045,
  "uuid": "[REDACTED]",
  "ttft_stream_ms": 1556,
  "time_to_request_ms": 229,
  "first_content_frame_ms": 1557,
  "queued_turn_count": 0,
  "result_index": 0
 }
]

Get new runs and ranking changes weekly