Comparison 04 of 10 · Markets

Rates, set beside FRED and Freddie Mac's own survey page

A desk that reads the New York Fed, the Treasury and Freddie Mac and labels the one number it has to infer. Its figures matched both publishers on the day; its history is slower to arrive.

Under test: labs.llc/rates/
Measured: 27 Sep 2026, 20:00–20:03 UTC, local copy of build 537
Rivals: FRED, series T10Y2Y · Freddie Mac PMMS

1 labs.llc ahead
3 a rival ahead
3 level
1 unconfirmed

1 What it is
2 How it works
3 The test
4 The checklist
5 Shortcomings
6 Verdict

1What it is, and who it is for

Rates is a desk for what money costs, read from the people who publish it: the Federal Open Market Committee's target range and when it last moved; the New York Fed's five overnight reference rates with percentile bands and volumes, plus the SOFR averages and index; the Treasury par yield curve with the two-year/ten-year spread drawn back to 1990 and every inversion shaded; a prime rate; and Freddie Mac's weekly mortgage survey.

It is for borrowers, savers, students of markets and journalists who want the official figures on one screen with each number's source named — not a commentary.

The Rates desk on labs.llc: the headline “What money costs, and who says so.”, with paragraphs describing the policy range, the New York Fed's overnight rates, the Treasury curve and a prime rate worked out from the target, and small badges marking figures as fetched or derived. — Rates at 1280 × 800, captured from the local copy of labs.llc build 537 that this review measured.

2How it works

Most of the reading happens in your browser. It calls the New York Fed's rates API directly (all/latest.json and an EFFR history search; rates/assets/js/app.js:63, 1488, 1503) and walks the Treasury's daily par-yield CSV one calendar year per request from 1990, three years in flight at a time (64–65, 1432, 1442–1443). Every request is made with cache: 'no-store' (277).

Prime is not read from anywhere. It is the target range's upper bound plus 3.00 points, the constant is commented as “a convention, not a rule and not a published rate”, and the panel shows the arithmetic (68–73, 1188–1196). Only the mortgage survey needs a relay, because a browser cannot read Freddie Mac's file directly: rates/pmms-relay.php fetches the PMMS history CSV, keeps the last 300 weeks, caches for six hours, serves a copy marked stale for up to a week if Freddie Mac fails, and remembers a failure for five minutes. It takes no parameters and has one hard-coded upstream.

3The test

Measured on 27 September 2026 (UTC). The two endpoints the browser calls were fetched with curl, one year of Treasury data at a time, exactly as the code builds them.

UTC	What	What came back
20:00:05	The page	HTTP 200, 63,595 bytes.
20:00:15	Mortgage relay	HTTP 200, 12,338 bytes, 97 ms: 300 weeks from 2020-12-31 (30-year 2.67, 15-year 2.17) to 2026-09-24 (30-year 7.03, 15-year 6.42).
20:02:36	New York Fed, latest rates	HTTP 200, 1,427 bytes, 47 ms, six records: EFFR 3.88 (24 Sep) inside a 3.75–4.00 target range; OBFR 3.88; TGCR 3.86; BGCR 3.86; SOFR 3.88 on $2,990bn. The derived prime is therefore 4.00 + 3.00 = 7.00%.
20:02:43	Treasury years 2026, 1990 and 1989	2026: 185 rows, latest 25 Sep, 19.2 s. 1990: 250 rows, 17.2 s. 1989: HTTP 200 with an empty body, 19.3 s — the behaviour the code comment records (app.js:75–79).
≈20:07	Cross-check with the publishers	The Treasury's 25 Sep row gives 2-year 4.81 and 10-year 5.17, a spread of 0.36; FRED's T10Y2Y page shows 0.36 for that day. Freddie Mac's page shows 7.03% and 6.42% for 24 Sep, as the relay did.

4The checklist

Eight checks. The two figure rows are where a desk like this is judged first, and it agreed to the hundredth. The rest split: Rates leads on putting everything on one screen, the publishers lead on history and on ways to take the data away.

Yes: does it
Partly: partly — read the note
No: does not
Not checked: not checked: no claim either way
Not applicable: does not apply

The checklist · Rates and 2 rivals8 checks

Marks for the rivals come only from their own pages, fetched on 27 September 2026 (about 20:07 UTC). A “?” is a question we could not settle, not a “no”.
Check	labs.llc · under testRates	rivalFRED · T10Y2Y	rivalFreddie Mac PMMS
2s10s spread for 25 Sep 2026Level	0.36	0.36	Not applicable
30-year mortgage, week of 24 Sep 2026Level	7.03%	Not applicable	7.03%
Policy rate, overnight rates, curve and mortgages on one pagelabs.llc ahead	Yes	Partly1	No2
Every 2s10s inversion since 1990 shadedUnconfirmed	Yes3	Not checked	Not applicable
Full history in one requestRival ahead	No4	Yes	Yes5
Mortgage history on offerRival ahead	300 weeks	Not applicable	since 1971
Downloads, an API or alertsRival ahead	No	Yes6	Partly7
Says how each figure was madeLevel	Yes8	Not checked	Yes9

1One series on this page; FRED offers data lists, which we did not test.
2Freddie Mac's own survey only, by design.
3Including one-day inversions (app.js:393, 445–457).
4One Treasury year per request, 17–19 s each here; re-fetched every visit.
5A download of rates since 1971.
6Download, the FRED API, ALFRED vintages and email notifications.
7The 1971 download was confirmed; an API or alerts were not checked.
8Figures tagged by source field; prime marked derived, with the arithmetic.
9Methodology and FAQs on the page.

Does better

FRED · T10Y2Y

The whole spread history as one series in one request.
Downloads, an API, ALFRED vintages and email notifications.
It publishes the next release date (28 September, when fetched).

Does better

Freddie Mac PMMS

It is the authority for the figure, with methodology and FAQs.
History back to 1971 as a download.
The release time is stated: Thursdays at 12 p.m. ET.

Goes further

Rates

Sets the policy rate, SOFR, the curve and the mortgage survey side by side, each read from its publisher.
Labels prime as derived and prints the sum, rather than presenting a convention as a published rate.
Shades every inversion since 1990 on the drawing itself.
Keeps showing the mortgage survey through a Freddie Mac outage for up to a week, and says the copy is stale.

5Where Rates falls short

Slow to fill the long history. Thirty-seven Treasury years, three at a time, at 17–19 seconds each from here, and nothing is cached between visits. FRED hands over the same spread as one series.
Mortgage history stops at 300 weeks (from 31 December 2020), where Freddie Mac offers everything since 1971.
No forward view. It shows what rates are, not what markets expect. CME Group describes its FedWatch tool as doing that from fed funds futures; the tool page timed out for us, so we only cite the description.
Prime is an inference by convention — stated openly, but still not a published series.
No downloads, alerts or API of its own.

6The verdict, by the checklist

Choose Rates to see what money costs across the whole system on one screen, straight from the publishers; choose FRED or Freddie Mac when you need the long history in hand.

The two rows that matter most for a figures desk — does it agree with the publisher? — came out level, to the hundredth, on the day we checked. After that the checklist divides cleanly. Rates is the better screen; the publishers are the better archive. The cost of reading the Treasury raw is time, and the review measured it.

Try Rates on labs.llc