Correction, October 1, 2026. Retatrutide is not the most accurately dosed compound in our data. The 67 percent this article reported for it was a June 27, 2026 snapshot and is superseded. Re-measured on October 1, 2026 with every source pooled, the June method, 7,991 of 15,633 retatrutide label comparisons (51.1 percent) land within 10 percent of the printed label, seventh of the twelve compounds below. With Finnrick's results left out, as the retatrutide lab-report page counts them, it is 5,106 of 11,248 (45.4 percent), with 52.9 percent more than 10 percent over and 1.7 percent more than 10 percent under: eighth of the twelve.
Two things moved the figure. The corpus grew and changed: 85 percent of the label comparisons added to our records for these twelve compounds since June 28 are Freedom Diagnostics reports, almost all added through a bulk import of that lab's reports that began on July 4, 2026 (before then our records held six community-submitted Freedom Diagnostics reports). And there are now two counts. Our retatrutide and tirzepatide lab-report pages leave all Finnrick results out of label comparisons, because Finnrick's own headline divergence figure uses the vendor's batch claim instead of the label when the vendor published one. The comparison this article counts is the same for every source, Finnrick included: the tested amount against the amount printed on the label. For Finnrick results it differs from Finnrick's headline figure on 692 of the 11,532 comparisons in these twelve compounds (6.0 percent). Leaving Finnrick out is those pages' counting rule, not a correction of Finnrick's measurements. Ten of the other eleven compound pages linked below still show the pooled count, so both counts are listed.
| Compound | Within 10%, all sources | Within 10%, Finnrick excluded | Over 10%, Finnrick excluded | Under 10%, Finnrick excluded | Finnrick only: over / under 10% |
|---|---|---|---|---|---|
| GHK-Cu | 55.9% (2,159 of 3,865) | 58.0% (1,697 of 2,924) | 37.7% | 4.3% | 19.4% / 31.5% (of 941) |
| Cagrilintide | 54.7% (599 of 1,096) | 53.5% (483 of 902) | 41.9% | 4.5% | 28.9% / 11.3% (of 194) |
| PT-141 | 54.7% (545 of 997) | 52.8% (448 of 848) | 39.0% | 8.1% | 17.4% / 17.4% (of 149) |
| Ipamorelin | 56.0% (908 of 1,621) | 52.7% (621 of 1,179) | 39.0% | 8.3% | 17.2% / 17.9% (of 442) |
| Tirzepatide | 53.8% (5,311 of 9,879) | 52.5% (3,748 of 7,142) | 45.7% | 1.8% | 26.7% / 16.2% (of 2,737) |
| MOTS-c | 51.3% (1,695 of 3,307) | 50.9% (1,642 of 3,228) | 46.5% | 2.6% | 16.5% / 16.5% (of 79) |
| CJC-1295 | 46.8% (389 of 832) | 50.4% (251 of 498) | 44.4% | 5.2% | 24.0% / 34.7% (of 334) |
| Retatrutide | 51.1% (7,991 of 15,633) | 45.4% (5,106 of 11,248) | 52.9% | 1.7% | 15.9% / 18.3% (of 4,385) |
| TB-500 | 46.8% (826 of 1,765) | 45.1% (670 of 1,484) | 52.3% | 2.6% | 19.9% / 24.6% (of 281) |
| Tesamorelin | 45.7% (1,748 of 3,825) | 44.9% (1,397 of 3,114) | 51.9% | 3.2% | 24.6% / 26.0% (of 711) |
| BPC-157 | 38.7% (1,360 of 3,518) | 36.1% (942 of 2,612) | 60.6% | 3.4% | 31.7% / 22.2% (of 906) |
| Semaglutide | 38.5% (690 of 1,794) | 35.8% (509 of 1,421) | 61.7% | 2.5% | 38.3% / 13.1% (of 373) |
Which reports you count changes the answer to this article's question. Pooled, and with Finnrick left out, more results run over the label than under it in all twelve compounds; without Finnrick the count is 18,056 over to 1,030 under. Without Finnrick, 83.2 percent of the comparisons are Freedom Diagnostics reports and 11.0 percent Janoshik; for retatrutide, 86.8 percent are Freedom Diagnostics. In Finnrick's results alone the split is close to even, 2,521 over to 2,304 under, and under-fills match or outnumber over-fills in eight of the twelve compounds (right-hand column).
What does not carry over from June: the June under-fill shares, the 2.6 to 1 and 1.4 to 1 over-to-under ratios, the median depth of over-fills and under-fills, and the cleanest and messiest rankings further down are not comparable with these figures. In June we called the Finnrick-only cut closer to the truth. We withdraw that reading. Neither set is a random sample of the vials sold: sellers decide which of their reports to publish, and every Finnrick sample was chosen by whoever sent or bought it. These data show that the sources disagree, not which one is closer to what a buyer receives. The June sections are kept below, for the record, and labelled as the June 27, 2026 snapshot.
Source: TitrateLab's COA database, read on October 1, 2026 at 01:58 UTC. Published report entries with a tested amount and a printed label amount, grouped by compound the way each /coa compound page groups them (CJC-1295 counts only reports filed under that exact name). Entries are stored results, not unique certificates. Within 10 percent means from 10 percent under to 10 percent over the label, inclusive.
The fear is getting shorted: paying for 10mg, injecting 6. Pooled across sources, in all 12 compounds we checked, more vials run over the label than under.
The fear that sends a grey-market peptide buyer hunting for a scale and a certificate of analysis is getting shorted: paying for 10mg and injecting 6. Pooled across sources, the data says that fear is mostly backwards. By June 27, 2026 we had collected 22,588 certificates of analysis and pulled the 15,167 that report a measured dose against the label. In every compound we looked at, more of the vials that missed the label missed it by running over, not under. On October 1 that still holds pooled, but not in every lab’s results; see the correction above. We did not run these assays. Independent labs did. We collected them, parsed them, and counted.
- The label is a starting point, not a fact. On October 1, 2026, the share of label comparisons within 10 percent of the printed amount runs from 38.5 percent (semaglutide) to 56.0 percent (ipamorelin) with every source pooled, and from 35.8 percent (semaglutide) to 58.0 percent (GHK-Cu) with Finnrick's results left out.
- Misses run over more often than under, depending on the source. Pooled, and without Finnrick, over-fills outnumber under-fills in all twelve compounds. In Finnrick's results alone, under-fills match or outnumber over-fills in eight of the twelve.
- Over-fills are common and fairly shallow. When a result runs more than 10 percent over, the median is 16 to 22 percent above label on October 1, depending on the compound. A nuisance for a careful titration, and still a dose you did not plan.
- Under-fills are rarer, and not consistently deeper. In the June 27 snapshot under-fills tended to run deeper than over-fills, with median under-fills of 25 to 26 percent short on tesamorelin, ipamorelin and CJC-1295. On October 1 the median under-fill is deeper than the median over-fill in seven of the twelve compounds pooled and in four of the twelve without Finnrick. The median under-fill is 22.9 percent short for tesamorelin pooled (15.6 without Finnrick), 18.2 for ipamorelin (15.6) and 39.7 for CJC-1295 (28.8).
- The compound sets your odds, and the ranking moves. Retatrutide is 51.1 percent within 10 percent of label pooled and 45.4 percent without Finnrick. The June snapshot's 67 percent for retatrutide is superseded; see the correction above.
How we measured this, and the one caveat that matters
TitrateLab does not run assays. We collect certificates of analysis that third-party labs produce. Each certificate reports, among other things, how much of the labeled compound the lab actually found in the vial. We parse that figure, store it as a percent deviation from the printed label, and aggregate it. The June 27, 2026 snapshot was mostly Finnrick and Janoshik reports, with smaller volumes from MZ Biolabs and from community submissions: out of 22,588 certificates collected then, 15,167 carried a usable single-compound dose-versus-label number, and that is the set the June sections below count. Since July 2026 most new label comparisons are Freedom Diagnostics reports; the October 1 figures and their source mix are in the correction at the top.
Here is the caveat, stated up front rather than buried at the bottom: a certificate of analysis is only as honest as whoever decided to publish it. A vendor can pay for several tests and post the one that came back generous, so a corpus that mixes vendor-published certificates with independent results may tilt toward over-fill. Over-fill is the flattering miss. Short fills can quietly stay in private chats.
June 27, 2026 snapshot: all sources pooled vs Finnrick only
Over-to-under, pooled and in Finnrick’s results
On October 1, 2026 Finnrick’s results run about 1.1 : 1 (2,521 over, 2,304 under)
We can measure how much the answer depends on the source. In the June 27, 2026 snapshot, pooled across every source, over-filled vials outnumbered under-filled ones by about 2.6 to 1. Restricted to Finnrick’s results, the ratio fell to 1.4 to 1. On October 1, 2026 the gap is wider: without Finnrick it is about 17.5 to 1 (18,056 to 1,030), and in Finnrick’s results alone about 1.1 to 1 (2,521 to 2,304). In June we called the Finnrick figure closer to the truth. We no longer say that. Sellers decide which reports they publish, and Finnrick’s samples were chosen by whoever sent or bought them, so neither is a random sample of what buyers receive. We publish both numbers side by side; these data show that they disagree, not which is right.
Pooled, most misses are over-fills, the opposite of the fear
Walk the chart below, which pools every source, and the pattern holds without exception. In all twelve compounds, the green over-filled bar is longer than the amber under-filled one. That still held in the October 1, 2026 re-measure, pooled and with Finnrick left out; in Finnrick’s results alone it did not (see the correction at the top). The ranking does not hold either: in this June 27 snapshot retatrutide landed on label most often and CJC-1295 least, and the re-measure in the correction at the top reorders both.
June 27, 2026 snapshot, all sources pooled (superseded, see the correction at the top)
Which way the vial misses
Share of vials over the label (green, right) vs under the label (amber, left), as a percent of all measured vials. The number at right is the share landing within 10% of label. Ranked cleanest to messiest.
For a buyer, the direction of the error is not academic, because the label is what you do dose math against, whether in your head or in a peptide reconstitution calculator. A grey-market vial arrives as lyophilized powder marked, say, 10mg. You reconstitute it, you draw what the label says is 2.5mg, and you inject. If the vial is over-filled by 17 percent, you did not inject 2.5mg. You injected closer to 2.9mg, and you have no signal that anything is off. For a GLP-1 that you are titrating slowly to keep nausea and gut side effects manageable, a quiet 17 percent overshoot is a faster ramp than you agreed to. For a research peptide it is overage you cannot bank on, because the next vial may not repeat it.
Over-fill may be deliberate: a vendor who builds in overage ships a vial that tests well against its label. That is an inference; the certificates do not record why a vial was filled the way it was. Generous is not the same as accurate, and a dose you cannot predict is still a dose you cannot control.
How far the misses run
In the June 27, 2026 snapshot, with every source pooled, the two directions of error were not mirror images. The typical over-fill ranged from 15 percent on MOTS-c and tirzepatide to 21 percent on GHK-Cu, while the typical under-fill ranged from 13 percent short on cagrilintide to 26 percent short on tesamorelin. On October 1, 2026 the over-fill band is similar: when a result runs more than 10 percent over, the median is 16 to 22 percent above label. Deeper under-fills are no longer a general rule. The median under-fill is deeper than the median over-fill in seven of the twelve compounds pooled and in four of the twelve without Finnrick, where it ranges from 13.5 percent short (cagrilintide) to 28.8 percent short (CJC-1295).
June 27, 2026 snapshot, all sources pooled: median miss, by direction
In June, over-fills clustered shallow and under-fills ran deeper
Where the typical miss landed relative to the label across the twelve compounds in June. The October 1 medians are in the text above.
The hazard is not the average. It is the spread. A market that over-filled every vial by exactly 15 percent would be trivial to dose around: you would draw a little less and move on. What you actually face is a label that may be wrong by an unknown amount in an unknown direction; in the June snapshot the typical misses ran from about a quarter short to a fifth over. You cannot dose precisely against that, and a batch that runs from near-empty to over-label cannot be dosed safely at any single setting, because you do not know which vial is in your hand until a lab tells you.
The cleanest compound, and the two messiest, in the June snapshot
This section reports the June 27, 2026 snapshot. Its retatrutide figure is superseded; the October 1, 2026 shares are in the correction at the top.
The averages hide how much the compound matters. In June the gap between the best and worst compound in the set was more than 20 points of on-label share: retatrutide 67 percent, BPC-157 and CJC-1295 44 percent. On October 1, with Finnrick left out, the gap is 22 points, from GHK-Cu at 58.0 percent to semaglutide at 35.8 percent (pooled, 56.0 percent for ipamorelin to 38.5 percent for semaglutide).
Retatrutide
Cleanest in the June snapshot
CJC-1295
Messiest · worst short-fill share
Tesamorelin
Deepest under-fill
In June, retatrutide led the set. That lead did not survive the October re-measure (51.1 percent pooled, seventh of twelve; 45.4 percent without Finnrick, eighth), so we have withdrawn the explanation we gave for it, that more buyers and more testing pushed vendors to fill it accurately. In the same June snapshot, CJC-1295 was the opposite end: only 44 percent of its vials landed on label and 26 percent ran short by a median of 25 percent, the worst short-fill share in the data. Tesamorelin was the deepest cutter, a 26 percent median on the vials that came up short. BPC-157 landed on label only 44 percent of the time too, but its misses skewed over, so the failure mode there was overshoot rather than shortfall. On October 1 the answer depends on the source: with Finnrick left out, 5.2 percent of CJC-1295 and 3.2 percent of tesamorelin results are more than 10 percent short, while in Finnrick’s results it is 34.7 percent and 26.0 percent. The June short-fill shares should not be read as current.
What this means if you are the one dosing
- Assume the vial is off, in either direction. In most of the lab reports we collect the miss is an over-fill, but in Finnrick’s results under-fills are about as common. Treat the printed dose as the vendor’s claim, not as the amount going into your body.
- The danger is variance, not direction. A steady 15 percent over you could plan around. A miss of unknown size in an unknown direction you cannot. That is why a single certificate on a clean batch does not make the next batch clean.
- No compound is reliably on label. On the October 1, 2026 count without Finnrick, the best compound in this set lands within 10 percent of label 58.0 percent of the time and the worst 35.8 percent; pooled, the range is 38.5 to 56.0 percent. The ranking moved between June and October, so do not treat any compound as the accurate one.
- A purity number is not a dose. Ninety-nine percent pure says the molecule is right. It says nothing about how many milligrams are present. Read the quantity line, not just the purity headline. We took that gap apart in our piece on pharma standards versus the grey market.
- One cert proves one vial. Dose accuracy is a property of the batch, and the only batch that matters to you is yours.
On a grey-market vial, the dose on the label and the dose in the vial are two different numbers. You paid for the first one. Only the second one goes into your body, and the only way to learn it is to read a certificate from a lab that had no reason to flatter the seller. Look up the vendor and the batch you are about to buy in our COA corpus, and run any certificate you already hold through the free COA verify tool before you draw a dose. The label is a claim. The assay is the fact.
Research and education only, not medical advice and not a dosing protocol. TitrateLab does not test peptides; we collect, parse, and aggregate certificates of analysis produced by independent laboratories and submitted by the community. Figures marked June 27, 2026 are computed from 15,167 single-compound dose-versus-label certificates out of 22,588 collected by that date, pooling all sources; the Finnrick-only cut narrowed the June over-to-under ratio from about 2.6 to 1 to 1.4 to 1. Figures marked October 1, 2026 come from the re-measure in the correction at the top, which supersedes the June retatrutide figure and gives each share with every source pooled and with Finnrick’s results left out. See our methodology. If we have a figure wrong, the contact link is in the footer, and we correct the record.
