Prepare a charity's quarterly tax-relief claim on donations with reach-back declarations, cancellations, company gifts and donor benefits
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
Best model for this task so far: Claude Haiku 5.5 (effort default (no --effort flag) · thinking on) (success, $0.0058, 24s).
All runs (3)
100% pass rate across 3 eligible runs: 3 pass, 0 partial, 0 fail. Partial and failed runs are listed below with everything else.ⓘ
| Model | Cost | Time | Step-insⓘ | Autonomy | Outcome | Trustⓘ | Date | Compare |
|---|---|---|---|---|---|---|---|---|
| AClaude Opus 5.5default · thinking on | $0.11 | 17s | 0 | Fully autonomous | pass | Trust 80% | 2026-10-22 | |
| AClaude Haiku 5.5default · thinking on | $0.0058 | 24s | 0 | Fully autonomous | pass | Trust 80% | 2026-10-22 | |
| AClaude Sonnet 5.5default · thinking on | $0.06 | 13s | 0 | Fully autonomous | pass | Trust 80% | 2026-10-22 |
Embed this result
A live badge for this use case. It links back here and updates as new runs are published.
Get new runs and ranking changes weekly
Which models got better or cheaper on real tasks like this one.