# Finance Agent v2 ![Category](https://img.shields.io/badge/Category-Finance%20Agent-1f6feb) ![Metric](https://img.shields.io/badge/Metric-%25%20Success-2da44e) ![Saturation](https://img.shields.io/badge/Saturation-Low-d29922) ![Created](https://img.shields.io/badge/Created-2026-555) > **Quick answer:** Finance Agent v2 is a second-generation benchmark testing AI agents on multi-step financial task automation — workflows that combine data retrieval, calculation, and financial decision-making rather than single-turn Q&A. Kimi K3 scores 54.4% as of July 2026. ## At a Glance **What it tests:** An agent's ability to complete multi-step financial workflows end-to-end, combining data retrieval, computation, and financial reasoning. **Why it matters:** Financial task automation (e.g., reconciliation, reporting, analysis pipelines) is a high-value enterprise use case where agentic reliability matters more than single-turn accuracy. **Known limitations:** As a "v2" iteration, exact differences from any prior version and detailed task composition are not yet independently published outside its citation by Moonshot AI. ## What Finance Agent v2 Measures Finance Agent v2 evaluates an agent's ability to complete multi-step financial workflows — such as retrieving financial data, performing calculations, and producing analysis outputs — testing agentic reliability across the full task pipeline rather than isolated question-answering. ## Benchmark Specifications | Field | Value | |---|---| | Task category | Agent / finance automation | | Metric | % task success | | Saturation | Low | | Created by | Not yet independently documented | ## How Finance Agent v2 Is Scored Agents complete multi-step financial workflows, with outcomes verified against expected results, producing an aggregate % task success rate. ## State-of-the-Art Results | Rank | Model | Score | Source | Date | |---|---|---|---|---| | 1 | [Kimi K3](/models/moonshot/kimi-k3) | 54.4% | [Kimi K3 technical report](https://www.kimi.com/blog/kimi-k3) | 2026-07 | *Score sourced from Moonshot AI's Kimi K3 announcement, July 2026.* ## Finance Agent v2 on Benchgen > No Benchgen results yet — [be the first to run Finance Agent v2](/benchmarks/platform/finance-agent-v2). ## Finance Agent v2 vs Other Benchmarks | Benchmark | What it tests | Saturation | |---|---|---| | **Finance Agent v2** | Financial task automation workflows | Low | | [CorpFin v2](/benchmarks/platform/corpfin-v2) | Corporate finance analysis & modeling | Low | | [SpreadsheetBench 2](/benchmarks/platform/spreadsheetbench-2) | Spreadsheet manipulation & formula tasks | Low | | [SaaS-Bench](/benchmarks/platform/saas-bench) | SaaS product usage workflows | Low | ## Run Finance Agent v2 on Your Model Benchgen lets you run Finance Agent v2 against your own model, tracking financial workflow automation task success over time. ## Frequently Asked Questions

What is Finance Agent v2?

Finance Agent v2 is a benchmark testing AI agents on multi-step financial task automation workflows.

What does a good score look like on Finance Agent v2?

Kimi K3 reports 54.4% as of July 2026, a solid frontier-level result reflecting meaningful but incomplete reliability in end-to-end financial automation.

Who created Finance Agent v2?

Finance Agent v2's originating team is not yet independently documented outside of its citation in Kimi K3's July 2026 technical report.

Rank	Model	Score
1	gemini-4-argon	65.4
2	kimi-k3	54.4

Finance Agent v2 — Results

Finance Agent v2 — Evaluations