EvasionBench
- Tests
- Detection and handling of evasive answers in corporate earnings calls.
- Best for
- Evaluating financial-analysis systems that must distinguish disclosure from deflection.
- Classification
- benchmark · emerging · open
AI evaluation domain · 9 source-backed records
Finance AI Benchmarks
Finance benchmarks span question answering, calculations, filing analysis, extraction, and professional workflows. Evidence quality and freshness matter as much as final-answer accuracy.
Decision guidance
What makes a finance AI benchmark useful?
Prefer tasks with traceable source documents, dated data, deterministic calculations, and explicit abstention rules. Re-test with the filings, chart of accounts, policies, and approval boundaries used in your own workflow.
9
Benchmarks in this domain
Sources first · alphabetical collection order
Runnability describes access to a usable repository, dataset, or harness—not whether setup is easy.
2026-07-15
FinanceAgent / FAB v2
- Tests
- Financial-agent benchmark for realistic analyst and finance workflow tasks.
- Best for
- Finance agents and analyst workflows
- Classification
- agent benchmark · current · commercial
FinanceBench
- Tests
- Open-book financial QA over public company filings with evidence strings.
- Best for
- Financial analyst and SEC filing workflows
- Classification
- benchmark · current · open
FinBen / PIXIU
- Tests
- Broad financial benchmark suite across extraction, QA, generation, forecasting, and decision-making.
- Best for
- Finance model evaluation suites
- Classification
- benchmark · current · open
FinQA / ConvFinQA
- Tests
- Financial numerical reasoning over reports using structured and unstructured evidence.
- Best for
- Financial math and reasoning
- Classification
- benchmark · current · open
GDPval
- Tests
- Economically valuable professional tasks across domains including finance, insurance, and operations.
- Best for
- Professional work automation signal
- Classification
- benchmark · emerging · mixed
Meta-Benchmarks for Financial Services
- Tests
- How 452 reported benchmarks map into 41 work activities and 38 banking business domains.
- Best for
- Designing evidence-weighted model selection for financial-services workflows.
- Classification
- toolkit · emerging · open
Open FinLLM Leaderboard
- Tests
- Open leaderboard for financial LLM evaluation.
- Best for
- Financial model comparison
- Classification
- scoreboard · current · hosted
QFBench
- Tests
- State-aware quantitative-finance agent tasks requiring code, market logic, and financial reasoning.
- Best for
- Quant finance agents
- Classification
- agent benchmark · emerging · open