11 records · Browse by task

Finance benchmarks

Find tests for financial research, document analysis, accounting, and tax work.

These tests evaluate specific tasks and agent setups. They do not establish investment performance or a single best model for finance.

Records keep metrics and harness variants separate. Source rows can include different model and agent configurations; linked profiles count distinct public model identities. Neither count represents independent evaluations. Dates describe the stored source record.

Public model rows count finite, displayable results linked to reviewed original artifacts. Publisher comparison charts can include other providers' models. Follow the result link to inspect each row's source and run conditions.

Financial research and reasoning

Finance Agent v2
78 stored source rows · 65 linked model profiles
Source update in snapshot: 2026-10-12
Source snapshot captured: 2026-10-15T18:47:28.403Z
Original source
The default Vals tool harness, partial-credit score, and all-pass score stay distinct. Version 1.1 appears below as historical evidence.
Financial analyst task categories
PRBench Finance
37 stored source rows · 19 linked model profiles
Snapshot label: October 15, 2026 snapshot
Original source
Published professional-reasoning configurations remain separate from finance-agent runs.
37 published rows
FinanceArena
19 stored source rows · 14 linked model profiles
Snapshot label: October 12, 2026 snapshot
Original source
The source table is labeled FinanceQA, Assumption-Based. It does not publish a task count or per-row run settings.
Professional financial-analysis questions

Financial documents and modeling

CorpFin v2
134 stored source rows · 77 linked model profiles
Source update in snapshot: 2026-08-12
Source snapshot captured: 2026-10-15T18:47:28.403Z
Original source
Credit-agreement results preserve exact-page and context-budget views.
Credit-agreement understanding tasks
MortgageTax
98 stored source rows · 58 linked model profiles
Source update in snapshot: 2026-09-01
Source snapshot captured: 2026-10-15T18:47:28.403Z
Original source
Semantic extraction and numerical extraction remain separate task views.
Mortgage and tax extraction tasks
EMB
74 stored source rows · 62 linked model profiles
Source update in snapshot: 2026-10-12
Source snapshot captured: 2026-10-15T18:47:28.403Z
Original source
Spreadsheet-modeling evidence comes from the published agent setup.
Excel-based financial modeling tasks

Accounting, banking, and tax workflows

Ramp Accounting Bench
16 stored source rows · 16 linked model profiles
Snapshot label: October 14, 2026 snapshot
Original source
A fixed accounting-agent setup reports mean criterion-level scores across three attempts.
137 accounting tasks · 3 attempts per task
BankerToolBench
2 public model rows linked to original artifacts
Original source
Sparse provider-reported banking-tool results are a starting point for inspection.
Finance and banking tool-use tasks
Tax Agent Bench
69 stored source rows · 60 linked model profiles
Source update in snapshot: 2026-10-12
Source snapshot captured: 2026-10-15T18:47:28.403Z
Original source
Overall and all-pass results use a dedicated US tax harness.
US tax research, calculations, forms, filings, and precedent analysis
TaxEval v2
145 stored source rows · 82 linked model profiles
Source update in snapshot: 2026-09-01
Source snapshot captured: 2026-10-15T18:47:28.403Z
Original source
Read tax-task conditions alongside each result.
Tax question answering and response evaluation

Historical evidence

Finance Agent v1.1
51 stored source rows · 31 linked model profiles
Archived source board
Source update in snapshot: 2026-06-04
Source snapshot captured: 2026-10-15T18:47:28.403Z
Original source
Archived Finance Agent v1.1 has a different task taxonomy and grading protocol from v2.
11 financial analyst task views

Other collections

All benchmarks