CyberMax-tools/figurewell-sec-xbrl-financials

Figurewell: SEC XBRL financial statements and key metrics

Need it fresh, filtered or via API? This free file is a snapshot (SEC XBRL financials up to the last refresh), last updated 2026-10-08.

• Insidewell on Apify ($0.004 per insider transaction): pulls today's SEC Form 4 trades for your own watchlist, filtered by buy/sell and size, with cluster-buy alerts on a schedule.
• Using it at work? Commercial license + support (from $49/year): invoice, PDF licence certificate, named support, freshness and availability commitments and an SLA; the data stays free and open for everyone.
• Get an email when this dataset updates: free, double opt-in, unsubscribe any time.
• CyberMax Store: every CyberMax data product, API plan and weekly brief in one place.

Information only, not investment advice.

The numbers from every financial statement filed with the SEC in the latest four quarters: 10-K, 10-Q, 20-F and 40-F reports from 7,000+ US-listed and SEC-registered companies, as tidy parquet you can query in one line. Revenue, net income, EPS, total assets, cash, debt, operating cash flow, capex and dividends per filing, plus every other reported face-statement value.

This build (see meta.json): 26,703 filings from 7,293 companies, filed 2025-10-01 to 2026-09-30 (SEC quarters 2025q4–2026q3): 16,834 quarterly reports (10-Q), 5,529 annual reports (10-K), 1,073 20-F and 147 40-F. 23,980 filings in key_metrics (21,679 with a stock ticker) and 5.9 million facts.

Why this and not the SEC API? EDGAR's companyfacts API is one company per request and returns every tag ever filed, in nested JSON. Screening, backtests, dashboards and LLM agents need "every company's latest revenue and cash" in one table. The public S&P 500 financial sets on Hugging Face stopped updating in 2023; this one covers every filer and refreshes each quarter.

Use it

from datasets import load_dataset
km = load_dataset("CyberMax-tools/figurewell-sec-xbrl-financials", "key_metrics", split="train").to_pandas()

# Latest annual numbers per ticker
annual = km[km.form == "10-K"].sort_values("accepted_utc").groupby("ticker").tail(1)
print(annual[["ticker", "company_name", "period_end", "revenue", "net_income", "net_margin", "free_cash_flow_ytd"]].head(20))

# Profitable small caps with net cash
q = km[(km.currency == "USD") & (km.net_income > 0) & (km.cash > km.long_term_debt.fillna(0)) & (km.total_assets < 1e9)]

DuckDB, straight from the Hub:

SELECT ticker, company_name, fiscal_period, revenue, net_income, eps_diluted
FROM 'hf://datasets/CyberMax-tools/figurewell-sec-xbrl-financials/data/key_metrics.parquet'
WHERE ticker IN ('AAPL', 'MSFT', 'NVDA') ORDER BY accepted_utc DESC;

Configs and columns

key_metrics (default, one row per 10-K/10-Q/20-F/40-F filing, amendments included): adsh (accession number), cik, ticker, company_name, form, fiscal_year, fiscal_period (FY, Q1–Q3), period_end, filed_date, accepted_et / accepted_utc (when EDGAR accepted the filing; see point-in-time below), currency (reporting currency; foreign filers report in their own, e.g. TWD, EUR), then the metrics:

Column	Meaning	Period
revenue, cost_of_revenue, gross_profit, operating_income, net_income	Income statement	The filing's own period: 12 months for FY, 3 months for a quarter
eps_basic, eps_diluted, diluted_shares	Per-share and weighted share count	Same as above
total_assets, current_assets, total_liabilities, current_liabilities, stockholders_equity, cash, long_term_debt	Balance sheet	At period_end
operating_cash_flow_ytd, capex_ytd, dividends_paid_ytd, free_cash_flow_ytd	Cash flow, as companies report it: fiscal year-to-date	cash_flow_quarters says how many quarters (1–4)
gross_margin, net_margin	Ratios to revenue	When revenue > 0

Plus sic (industry code), filer_status (1-LAF large accelerated … 5-SML smaller reporting), business_country, business_state, amended_later (a later amendment replaced this filing), filing_url, tags_used (JSON: which XBRL tag each metric came from, e.g. RevenueFromContractWithCustomerExcludingAssessedTax) and source_quarter.

Metrics use each company's own reported value for standard US-GAAP or IFRS tags, trying common alternatives in a fixed order (see tags_used), with no segment breakdowns. Nothing is estimated or adjusted: a metric is null when the company didn't report it that way (banks have no gross profit, many companies report no long-term debt line).

facts (5.9M rows): every numeric value on the financial statements without segment/axis breakdowns: adsh, cik, ticker, company_name, form, fiscal_year, fiscal_period, filed_date, accepted_utc, taxonomy (us-gaap, ifrs, dei, srt, or custom for company-specific tags), tag, period_end, duration_quarters (0 = point in time), uom, value, source_quarter. Prior-period comparatives in each filing are included (filter on period_end).

filings (26,703 rows): one row per filing, including registration statements (S-1, S-4, F-1…) and 11-K plan reports that carry XBRL, with company address fields, accepted_et / accepted_utc and filing_url.

Values are exactly as filed (units, not thousands). period_end is the SEC's month-end rounding of the fiscal period end (Apple's quarter ended 28 March shows as 2026-03-31). Per-share values are in currency units. If a company restated a number later, both filings are here: for today's view take the latest accepted_utc; for backtests see point-in-time below.

Backtests: use the value as known at the time (point-in-time)

Every row carries accepted_utc (and accepted_et): the moment EDGAR accepted the filing, from the accepted field of the SEC's SUB table. It is US Eastern time to the minute (the SEC data set drops the seconds: Apple's FY2025 10-K reads 2025-10-31T06:01:00-04:00 here and 06:01:26 on its EDGAR index page), so treat it as "public by the end of that minute". Use it, not filed_date, to decide when a number became public:

• Intraday timing. filed_date is the EDGAR filing date. Under Regulation S-T a filing accepted up to 5:30 pm ET still gets that day's date, so a 10-Q accepted at 4:15 pm ET has the same filed_date as one accepted at 7 am. Joining on filed_date and trading that day's close uses numbers released after the close. Compare accepted_et with your market's session times (and your own holiday calendar) and act from the next session when it was accepted after the open.
• Restatements. When a company files an amendment (10-K/A, 10-Q/A) or restates a prior period in a later filing, both versions are here. For "what's true now", take the latest accepted_utc. For a backtest decision at time T, take the latest value accepted on or before T; taking the latest overall is look-ahead.

import pandas as pd
km = load_dataset("CyberMax-tools/figurewell-sec-xbrl-financials", "key_metrics", split="train").to_pandas()

def as_of(km, when):
    """Each company's latest filing per fiscal period that was public at `when` (UTC)."""
    t = pd.Timestamp(when, tz="UTC")
    known = km[km.accepted_utc <= t]
    return known.sort_values("accepted_utc").groupby(["cik", "fiscal_year", "fiscal_period"]).tail(1)

snap = as_of(km, "2026-05-01 20:00")   # 4 pm ET on 1 May 2026: what a strategy could have known then

Same rule in DuckDB, for one metric of any period, including restated prior-period comparatives (the facts config):

SELECT cik, ticker, tag, period_end, duration_quarters,
       arg_max(value, accepted_utc) AS value_known, max(accepted_utc) AS known_since
FROM 'hf://datasets/CyberMax-tools/figurewell-sec-xbrl-financials/data/facts.parquet'
WHERE tag = 'Revenues' AND accepted_utc <= TIMESTAMPTZ '2026-05-01 20:00:00+00'
GROUP BY ALL;

The window is the latest 4 SEC quarters, so a restatement filed before the window starts is not visible here. Information only, not investment advice.

Refresh

Rebuilt each quarter from the SEC's Financial Statement Data Sets a few days after they are published (next: 2026q4, in January 2027). meta.json has the build time, quarters and coverage counts per metric.

Changes

• 8 Oct 2026: added accepted_et / accepted_utc to every config and the point-in-time section (thanks to a user report in discussion #1).

Licence

SEC EDGAR filings and the Financial Statement Data Sets are US government public records, free of copyright restrictions; this compilation is released under CC0 1.0. Compiled by CyberMax. Not affiliated with or endorsed by the SEC. Not investment advice; check filing_url before relying on a number.

Also from CyberMax

• Insidewell: SEC Form 4 insider trading tracker: insider buys and sells for any watchlist of tickers, with cluster-buy flags, on demand or on a schedule. Screen on fundamentals here, then watch who's buying.
• Free dataset: Riskroll: 10-K/10-Q Risk Factors and MD&A text, the words behind these numbers.
• Free dataset: Eventwren: SEC 8-K material events by Item (earnings, CEO/CFO changes, restatements).
• Free dataset: SEC Form 4 insider transactions, latest 5 EDGAR days.
• More tools for AI agents and data teams: https://apify.com/cybermax

More free from CyberMax

• All free CyberMax demos & datasets
• Riskroll: SEC 10-K & 10-Q Risk Factors, MD&A and Market Risk Text (clean, by section)
• Eventwren: SEC 8-K Material Events by Item (latest EDGAR filings)
• Insidewell: SEC Form 4 Insider Transactions (latest EDGAR filings)

Quickstart notebook

Load this dataset with pandas and see real outputs in the quickstart notebook.