Finch (FinWorkBench)
Mistral Large 4 (Le Chonk, open weights) 73.3% In use Agent completes 172 enterprise finance and accounting spreadsheet workflows (384 tasks) from real data. Human plus calibrated LLM-judge grading.
ACL 2026 Findings.
In the authors’ words
We introduce FinWorkBench (a.k.a. Finch) for evaluating AI agents on real-world, enterprise-grade finance and accounting workflows that interleave data entry, structuring, formatting, web search, cross-file retrieval, calculation, modeling, validation, translation, visualization, and reporting. Finch is sourced from authentic enterprise workspaces from Enron (15,000 files and 500,000 emails) and other financial institutions, covering the period 2000–2025 and preserving the in-the-wild messiness of multimodal artifacts such as tables and charts across diverse domains including budgeting, trading, asset management, and operational management. We propose a workflow construction process that combines LLM-assisted mining of workflows from authentic enterprise environments with expert annotation: (1) LLM-assisted, expert-verified derivation of workflows from real-world email threads and spreadsheet version histories, and (2) meticulous annotation requiring over 700 hours of expert effort. This yields 172 composite workflows with 384 tasks, involving 1,710 spreadsheets with 27 million cells, along with PDFs and other artifacts, capturing the intrinsically messy, long-horizon, knowledge-intensive, and collaborative nature of real-world enterprise work. We conduct both human and automated evaluations of frontier AI systems, including GPT-5.1, Claude Sonnet 4.5, Claude Opus 4.5, Gemini 3 Pro, Grok 4, and Qwen 3 Max. Under human evaluation, GPT-5.1 Pro spends an average of 16.8 minutes per workflow yet passes only 38.4% of workflows. Comprehensive case studies further surface the challenges that real-world enterprise workflows pose for AI agents.
- Level
- Expert Specialists with years of training, often over hours of work
- Scored
- Graded by people Human experts grade or compare the outputs.
- Answers
- Open-ended Free-form text, documents, designs or media judged as a whole.
Mistral Large 4 (Le Chonk, open weights) leads with 73.3%
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Mistral Large 4 (Le Chonk, open weights) | 73.3% | Oct 2026 | Model card |
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
One lab put it in a launch post in the last six months
Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.
| Date | Launch post | Score | As the post writes it |
|---|---|---|---|
| 29 Oct 2026 | Mistral Large 4 (Le Chonk, open weights) | 73.3 | FinWorkBench (Finch) |
More in Law, finance and medicine
- HealthBench Professional In use 525 physician-authored tasks from real clinician chats (care consult, documentation, research) with physician-adjudicated rubrics. Claude Sonnet 5.5 69.2%
- MedXpertQA-MM In use Models answer expert-level medical exam questions that combine images with clinical records and test results. Scoring is accuracy.
Qwen 3.8 Max 80.4% - HealthBench In use 5,000 realistic health conversations graded against physician-written rubrics. A length-adjusted variant penalizes verbosity. GPT-5 67.2%
- KBL In use Models answer Korean legal knowledge and reasoning tasks and Korean bar exam questions. It has 7 knowledge tasks, 4 reasoning tasks and 2,510 bar exam items.
Solar Pro 4 77.5% - KorMedMCQA In use Models answer 7,469 multiple-choice questions from Korean licensing exams for doctors, nurses, pharmacists and dentists, held from 2012 to 2024. Solar Pro 4 93.2%
- PRBench In use Model answers 1,100 expert-written law and finance questions. LLM judge grades against weighted expert rubrics of 10-30 criteria each.
Muse Spark 1.3 61.6%