DAYJOB: Tax
Surge AI's benchmark of 50 end-to-end professional tax assignments; the best published model passes 10.8% at a 95% rubric threshold.
Overview
Surge AI published DAYJOB: Tax on October 27, 2026, the third benchmark in its DAYJOB family after Healthcare and Finance. It has 50 environments across eight practice areas (individual, small business, partnerships, property and investments, estates/gifts/trusts, international, corporate transactions, review and disputes) and 1,789 files, mostly PDFs plus Excel, Word and images such as handwritten receipts. Agents must produce returns, workpapers, technical reviews and client advice, find issues clients did not raise and recognize insufficient information. On pass@1 at a 95% rubric threshold, Claude Sonnet 5.5 scores 10.8%, Claude Fable 5.1 9.6% and Claude Opus 5.5 9.2%.
This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.
Key features
- 50 environments across eight tax practice areas
- Median 30 files per task, up to 123; 1,789 files in total
- Pass@1 at a 95% rubric threshold: Sonnet 5.5 10.8%, Fable 5.1 9.6%, Opus 5.5 9.2%
- Dataset on Hugging Face (surgeai/DAYJOB-tax); full access by request
Use cases
- Measuring whether agents can complete professional tax work end to end
- Comparing models on cost versus performance for document-heavy workflows
Pricing
Public leaderboard and Hugging Face dataset; full benchmark access by request to Surge AI, pricing unknown.
Pricing reflects the source record and may change. Confirm current terms with the provider.
Dates and sources
- Released
- 27 Oct 2026
- Record updated
- 29 Oct 2026
- Record last checked
- 29 Oct 2026