Press | Montana Research Foundation

Media contact

Interviews, embargoed copies of upcoming reports, and quotes go through Francieli Carra. The form goes straight to the team and is labelled as a press request.

Send a press request

Fast facts

Status: Independent non-profit research institute. No investors.
Research areas: Foundations of AI · Trustworthy Systems · Applied Research · Open Science
Preprints: 4 published, 4 with a DOI on Zenodo.
Insight reports: 3 published, each with a gated PDF and a full web version.
Open data: github.com/Montana-Research-Foundation/open-data, run records and a verification script per study.
Site: montanaresearch.org

Boilerplate

Use either version as written. Attribution is "Montana Research Foundation".

Short · about 50 words

Montana Research Foundation is an independent non-profit research institute doing open research on AI agents. Its researchers publish preprints with DOIs and insight reports on agent evaluation, coding agents, and agents in financial services. Every reported number regenerates from public run records. More at montanaresearch.org.

Long · about 100 words

Montana Research Foundation is an independent non-profit research institute focused exclusively on open AI research. Its PhD-level researchers work on foundations of ai, trustworthy systems, applied research, open science. The foundation publishes preprints with DOIs on Zenodo and a series of insight reports for lab, industry, and policy readers, covering agent evaluation, coding agents in production, and agents in financial services. Run records, task families, and verification scripts for every study are public in the foundation's open-data repository, so each figure can be regenerated by anyone. The foundation has no investors and no product roadmap. More at montanaresearch.org.

Reports and charts

Every chart from every report, as SVG and PNG, licensed CC BY 4.0. Each file carries its own caption, source line, and licence, so it can be embedded as is. Credit Montana Research Foundation and link to the report.

MRF-R-2026-03 · August 2026 · Dr. Ricardo Arcifa, Francieli Carra

Agents in Financial Services

Adoption, customer-facing risk, model error rates, and the rules already in force

01 AI adoption in UK financial services, 2024 SVG PNG
02 US consumers who have interacted with a bank chatbot SVG PNG
03 Hallucination rate on verifiable questions about US federal cases SVG PNG

MRF-R-2026-02 · August 2026 · Dr. Ricardo Arcifa, Francieli Carra

Coding Agents in Production

What controlled trials, delivery data, and the benchmarks themselves say about AI-assisted software engineering

01 METR 2025: expected speed-up against measured change in task time SVG PNG
02 Stack Overflow 2025: developers' trust in the accuracy of AI tools SVG PNG
03 SWE-Lancer: value earned out of $1 million of real freelance tasks SVG PNG
04 HumanEval: cost of one evaluation run by agent architecture SVG PNG

MRF-R-2026-01 · August 2026 · Dr. Ricardo Arcifa, Francieli Carra

The State of Agent Evaluation

What certified task families reveal about how frontier labs measure agents

01 Defective fixtures refused, by pipeline stage SVG PNG
02 Pass rate per family and configuration, 20 seeds per cell SVG PNG
03 Opus 5: declared confidence vs measured generalisation SVG PNG
04 GPT-5.6 Sol: declared confidence vs measured generalisation SVG PNG
05 Opus 5: mean reasoning tokens per rollout, confirmatory stage SVG PNG
06 Outcomes by label, confirmatory stage, 30 seeds per arm SVG PNG

Preprints

Logos

Vector files for print and web. Keep the marks unaltered and give them clear space.

Icon, light on dark For dark grounds. Download
Icon, dark on light For light grounds. Download
Icon, purple Brand accent, vector. Download
Icon, purple (PNG) Raster fallback. Download
Wordmark, light on dark Full lockup, vector. PNG: download Download
Wordmark, dark on light Full lockup, vector. PNG: download Download

Press and media