Numa 1.

Built for

financial research.

Numa 1 is our most capable financial AI engine. It analyses the filings, works out the numbers itself and shows the source behind every figure.

Numa 1 works every figure out from the filings and links each one to the page that backs it, so you can check any number in a click.

Does the research of an analyst, in seconds

It analyses the whole filing

Every page, not a summary of it. Ask what changed since the last one and it tells you: new numbers, new plans, guidance that moved.

It compares like an analyst

Valuation, growth and margins against the companies that matter, worked out from the reported figures.

It stress-tests a portfolio

Where you are concentrated, and how what you hold now would have fared in 2008, 2020 and 2022.

It argues both sides

Both sides of the story for any stock, drawn from the record, so you decide with the full picture.

Benchmarks

Table of four tests comparing Numa 1 by Moneda, with tools, against GLM-5.2 by Z.ai, with no tools, and the best other model with a score for each test; the finance knowledge row is with no tools for all three. Groundedness, meaning figures that come from data and are not made up: 97.9% against 13 to 17%. Claim verifiability, meaning claims supported by their cited source, which is one of the three dimensions of JPMorganChase's Deep FinResearch Bench (arXiv 2604.21006), as implemented by Moneda: 92% against 3%. Expert finance tasks from Vals AI's Finance Agent Benchmark v2: 80% on the public set, run and graded by Moneda, against 8%, and 65.4% for Gemini 4 Argon, the top score on Vals AI's own leaderboard, measured on its private set. Finance knowledge on the FLARE-CFA questions: 95.6% for Numa 1, 95.5% for GLM-5.2, and 95.9% for Opus 4.8.

October 2026.

1 Groundedness: whether a figure in an answer comes from data or is made up. Questions with a known answer.

2 Claim verifiability: claims supported by their cited source. One of the three dimensions of JPMorganChase's Deep FinResearch Bench (arXiv 2604.21006). Run by Moneda.

3 Expert finance tasks: share of credit earned on the public set of Vals AI's Finance Agent Benchmark v2. Run by Moneda. The 65.4% is Vals AI's top published score, on its private set.

4 Finance knowledge: FLARE-CFA, practice questions in the style of the CFA exam, not the exam itself.

Research you can rely on

No made-up numbers

Figures come from market data and filings, never from memory.

Sources you can open

Each claim links to the document that backs it.

It knows what it knows

When the data is not there, it says so. It does not guess.

You stay in charge

It researches and prepares. Nothing happens outside what you have approved.

FAQs

What is Numa 1?

Numa 1 is Moneda's financial AI engine. It analyses company filings, works out the numbers and shows the source behind each figure.

Is Numa 1 the same as Moneda AI?

No. Moneda AI is the agent in the app. Numa 1 is the financial AI engine that powers it.

How is Numa 1 different from a general AI model?

A general model answers from what it remembers, and it can be confidently wrong. Numa 1 works from filings and market data and shows where each figure came from, so you can check it.

What can I ask Numa 1?

What changed in a company's last filing, how a stock compares with its peers, the bull and the bear case, or how your own portfolio would have held up in past crises.

Does Numa 1 give investment advice?

No. It gives information and research. Every decision to buy or sell is yours.

How do I use Numa 1?

Open the Moneda app and ask Moneda AI a question. It runs on Numa 1 on every plan.