Claude Haiku 5.5 (effort default (no --effort flag) · thinking on) on Prepare a charity's quarterly tax-relief claim on donations with reach-back declarations, cancellations, company gifts and donor benefits

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

Trust 80%Fully autonomoussuccess · first tryCompare

Estimated cost: $0.0058
Wall time: 24s
Human step-ins: 0
Outcome: Pass

Compare with other models on this task →

Exact prompt

# Prepare a small charity's gift aid claim for the quarter Inputs: `donations.csv` (every donation received July to September 2026: donor, date, amount, donor type and the value of any benefit the donor received in return) and `declarations.csv` (gift aid declarations donors have signed, with any cancellation date). Every amount is in pounds. This is a made-up scheme for this task; follow these rules, not any real tax rules. Claim rules: - A donation is covered only if its donor has a declaration whose declaration_date is no more than 4 years after the donation date (declarations reach back up to 4 years, and cover everything after), and, if that declaration has a cancelled_date, the donation was made before the cancelled_date. - Donations from companies (donor_type 'company') are never eligible. - A donation where the donor received a benefit is eligible only if benefit_value is no more than 25% of the amount; otherwise none of it is eligible. - Gift aid = 25% of each donor's total eligible donations, rounded half-up to the cent per donor. Write `out/claim.json`: `{"covered": {"<donation_id>": true|false, ...} for every donation from an individual (whether a declaration covers it, ignoring the benefit rule), "benefit_ok": {"<donation_id>": true|false, ...} for every donation from an individual with a benefit_value, "eligible_total": {"<donor_id>": <sum of that donor's eligible donations>, ...}, "gift_aid": {"<donor_id>": <gift aid>, ...}, "claim_total": <sum of the per-donor gift aid figures>}` listing only donors with at least one eligible donation. ## 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.

Work out a small charity's claim on a quarter's donations under a made-up relief scheme. Declarations reach back four years, a cancellation stops cover from its date, company gifts never qualify, a donation with a benefit over a quarter of its value is excluded, and the claim is 25% of each donor's eligible total rounded per donor. Ignoring reach-back, cancellations or benefits each change the claim. 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: declaration_reach_back cancellation benefit_limit companies_and_gift_aid

Setup notes: Batch 24 case z05-gift-aid-claim (offline, synthetic inputs, seed 896512830). 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 1.78 (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: $0.005799. 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: $0.0058
Wall time: 24s
Human step-ins: 0
Turns: 4

Model usage
Model	Input tok	Output tok	Cost
AClaude Haiku 5.5anthropic/claude-haiku-5.5effort default (no --effort flag) · thinking on	65,180	5,432	$0.0058

Reproduce this run

Use the use case prompt on the use case page.
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.
It counts as a success only if it meets every success criterion: declaration_reach_back cancellation benefit_limit companies_and_gift_aid
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).

Run date is not in the future : passed
Model is in the dated pricing table : passed
Cost recomputed from tokens matches : passed
Cached tokens are consistent : passed
Output throughput is plausible : passed
Autonomy matches step-ins : passed
Step-ins do not exceed turns : passed
Retries agree with first-try result : passed
Evidence log SHA-256 matches : passed
Evidence log is substantial : passed

Evidence log

[
 {
  "duration_api_ms": 22996,
  "stop_reason": "end_turn",
  "session_id": "[REDACTED]",
  "total_cost_usd": 0.00579901,
  "usage": {
   "input_tokens": 6,
   "cache_creation_input_tokens": 12793,
   "cache_read_input_tokens": 52381,
   "output_tokens": 5432,
   "output_tokens_details": {
    "thinking_tokens": 3400
   },
   "server_tool_use": {
    "web_search_requests": 0,
    "web_fetch_requests": 0
   },
   "service_tier": "standard",
   "cache_creation": {
    "ephemeral_1h_input_tokens": 12793,
    "ephemeral_5m_input_tokens": 0
   },
   "inference_geo": "global",
   "iterations": [
    {
     "input_tokens": 2,
     "output_tokens": 621,
     "cache_read_input_tokens": 20528,
     "cache_creation_input_tokens": 4804,
     "cache_creation": {
      "ephemeral_5m_input_tokens": 0,
      "ephemeral_1h_input_tokens": 4804
     },
     "type": "message"
    }
   ],
   "speed": "standard",
   "fallback_credit": null
  },
  "modelUsage": {
   "claude-haiku-5-5": {
    "inputTokens": 6,
    "outputTokens": 5432,
    "cacheReadInputTokens": 52381,
    "cacheCreationInputTokens": 12793,
    "webSearchRequests": 0,
    "costUSD": 0.00579901,
    "contextWindow": 1000000,
    "maxOutputTokens": 128000,
    "thinkingTokens": 3400,
    "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": 4,
  "subtype": "success",
  "api_error_status": null,
  "result": "I wrote the claim to `out/claim.json`. An independent recomputation from the CSVs matches every section.\n\n**Result:** 7 donors have eligible donations, and the claim total is \u00a3111.49.\n\n| Donor | Eligible total | Gift aid |\n|---|---|---|\n| DN303 | 50.00 | 12.50 |\n| DN312 | 155.95 | 38.99 (38.9875, rounded half-up) |\n| DN316 | 75.00 | 18.75 |\n| DN326 | 30.00 | 7.50 |\n| DN344 | 55.00 | 13.75 |\n| DN355 | 5.00 | 1.25 |\n| DN364 | 75.00 | 18.75 |\n\n**Choices I made:**\n- **Four-year rule:** every declaration is dated 2025 or 2026, so the \"no more than 4 years\" test is met for all donations. Donations made before a declaration (for example G9003 and G9026) are covered, because declarations reach back.\n- **Cancellation:** a donation counts only if it is dated before the donor's `cancelled_date`. This excludes G9008 and G9009 (DN316), and G9023, G9024 and G9025 (DN360).\n- **Benefit rule:** `benefit_ok` is recorded for every individual donation with a benefit value, whether or not it is covered. G9024 is exactly 25% and passes; G9027 is exactly 25% and passes; G9002, G9009, G9010 and G9019 fail.\n- **Companies:** DN336 is a company donor, so its donations are left out of `covered` and `benefit_ok`. Its declaration does not make them eligible.\n- **Donors with nothing eligible:** DN328, DN336 and DN360 are left out of `eligible_total` and `gift_aid`, as the spec requires.\n\nI did not write anything to `ask/` and did not use the network.",
  "ttft_ms": 1211,
  "type": "result",
  "duration_ms": 23311,
  "uuid": "[REDACTED]",
  "ttft_stream_ms": 1109,
  "time_to_request_ms": 225,
  "first_content_frame_ms": 1110,
  "queued_turn_count": 0,
  "result_index": 0
 }
]