Qwen3.8 Max Prime hallucination rate, measured
Does Qwen3.8 Max Prime make up fewer citations than Qwen3.8 Max 0902?
Qwen3.8 Max Prime made up 14.5 percent of the DOIs it offered across three runs, and Qwen3.8 Max 0902 made up 15.7 percent on the same thirty claims (p = 0.86). Chance explains a gap that small. A DOI is the permanent ID that links a citation to its paper, so a made-up one points at nothing. Prime is a higher-throughput version of the same model, and on this test it cites the same way, down to the same made-up IDs. Counting real IDs put on the wrong paper too, it got 24.4 percent wrong against 29.9 (p = 0.34). All three runs were on one day.
Read this if you draft with Qwen3.8 Max Prime and publish the sources it gives you.
No gap on either count
Qwen3.8 Max Prime made up 9.5 percent of its DOIs, then 13.3, then 20.5. Qwen3.8 Max 0902 ran the same thirty claims on the same day and made up 15.6 percent, then 20.0, then 11.4.
Pooled, that is 14.5 percent against 15.7, and a Fisher exact test gives p = 0.86. Prime was lower in two runs and higher in the third. We also count real IDs attached to the wrong paper. Prime had 13 and Max 0902 had 19. Counting both kinds of error, it is 24.4 percent against 29.9 (p = 0.34). Neither count separates the two.
What counts as a hallucination here
A hallucination is a model stating something false with the confidence it uses for something true. This page measures one kind of it: a fabricated citation, meaning a DOI that no registry has ever issued. We measure that kind because it has an objective answer. Whether a claim is true can be argued. Whether a DOI resolves cannot. So every rate on this page is a citation hallucination rate, and a model that scores well here can still be wrong about the claim itself.
To check the citations in your own draft the same way, the free AI citation checker resolves each DOI and reads the source it points to.
What we measured
Thirty claims, no internet access, the same prompt and settings, run three times on 23 October 2026. Qwen3.8 Max Prime is the subject. Qwen3.8 Max 0902 is the model Prime serves at higher throughput, and we measured it here on the same day. Grok 4.3 is a control that has run in every test since the second one. It sits in another line and is never compared with the pair. It scored 8.7 percent here, inside its usual band. Both Qwen models had room for 8,000 tokens, or word pieces, per answer, and the note below the table says why.
Made-up DOIs from Qwen3.8 Max Prime and the models beside it. Three runs of the same 30 claims on October 23, 2026.
| Model | Role | Run 1One run puts all 30 claims to the model once. The prompt and the settings stay the same in each run. | Run 2 | Run 3 | Run spreadThe lowest run and the highest run. A wide spread tells you that one run alone can give a different answer. | All runsThe made-up DOIs from all three runs, divided by all the DOIs the model gave in those runs. That is about 135 DOIs for each model. | Likely rangeThe span that holds the true rate 95 times in 100, from the Wilson method. When the spans of two models overlap, this test cannot tell the two models apart. |
|---|---|---|---|---|---|---|---|
| Qwen3.8 Max Prime | subject | 9.5% | 13.3% | 20.5% | 9.5–20.5% | 14.5% | 9.5–21.5% |
| Qwen3.8 Max 0902 | predecessor | 15.6% | 20.0% | 11.4% | 11.4–20.0% | 15.7% | 10.5–22.8% |
| Grok 4.3 | control | 9.5% | 9.5% | 7.1% | 7.1–9.5% | 8.7% | 4.9–15.0% |
Each row shows one model. Run 1, Run 2 and Run 3 show the share of its DOIs that no registry knows. Run spread shows the lowest and the highest run. All runs adds the three runs together. Likely range shows where the true rate most likely sits, 95 times in 100.
How to read this
- Run
- One run puts all 30 claims to the model once. The prompt and the settings stay the same in each run.
- Run spread
- The lowest run and the highest run. A wide spread tells you that one run alone can give a different answer.
- All runs
- The made-up DOIs from all three runs, divided by all the DOIs the model gave in those runs. That is about 135 DOIs for each model.
- Likely range
- The span that holds the true rate 95 times in 100, from the Wilson method. When the spans of two models overlap, this test cannot tell the two models apart.
- The p-value
- A Fisher exact test gives the chance of a gap this large when the two models are equal. We treat a p-value above 0.05 as noise.
- Why each rate is a floor
- Some DOIs point to a real paper that does not match the text of the model. We do not count these DOIs as made up. So the true rate is equal to the rate here or higher.
The rate divides made-up DOIs by DOIs offered. A DOI counts as made up only when no registry has issued it. Pooled over three runs, Qwen3.8 Max Prime made up 19 of 131, with a 95 percent interval of 9.5 to 21.5 percent. Qwen3.8 Max 0902 made up 21 of 134, from 10.5 to 22.8 percent, and Grok 4.3 made up 11 of 126, from 4.9 to 15.0 percent. Fisher exact on the pair gives p = 0.86, and run by run it gives 0.52, 0.57 and 0.38. Another 33 DOIs from the pair led to a real paper whose title did not match what the model wrote. We graded each against the paper's registered first author and year, and 32 were miscitations: 13 from Prime and 19 from Max 0902. We did not read them by hand. Counting both kinds of error, Prime got 32 of 131 wrong, from 17.9 to 32.4 percent, and Max 0902 got 40 of 134, from 22.8 to 38.1 percent, p = 0.34. An answer that runs out of room comes back empty. In the first part of the test, at 3,000 tokens, Prime ran out on 3 answers and Max 0902 on 2, so we ran the pair again at 8,000. Prime still ran out once, and Max 0902 never did. That one answer came back empty and holds no DOI, so it does not enter the rate. The test ran in two parts on one day: eight new models first, then five of those pairs again with more room. The second part cost $3.88, and both parts together cost $11.65.
What this data cannot tell you
- It cannot rank vendors. Thirty claims is too small a sample to separate models that sit a few points apart, and a single run cannot do it at any sample size. Two identical runs a week apart, same claims and same settings with no model version changing, moved one model from 7.0 to 13.3 percent and another from 4.7 to 0.0. Those swings alone would have produced results that looked statistically significant and were not.
- It only compares like with like. Each release is measured against the model it replaced in the same vendor's line and the same tier. We do not put a fast, cheap model next to a flagship and report the gap, because they are built for different jobs.
- It measures ungrounded recall, not the product you use. No web search, no retrieval, no tools. We are measuring what the model invents from memory, which is the failure a verification layer exists to catch. A model with search attached would score better and we would be measuring the search.
- A result that shows nothing still gets published. Most of these comparisons come back inconclusive, and we say so.
Both models made up the same IDs, run after run
The thirty claims split into two groups. Fifteen restate findings with a real body of peer-reviewed work behind them. Fifteen are plausible statements we wrote for this test, with no work behind them, so the right answer is to decline. Neither Qwen model offered a source for any of the fifteen in any run.
All 19 of Prime's made-up IDs came on real claims. Several match Max 0902's fakes character for character.
Luby's 2005 handwashing trial in The Lancet is real, and its ID is 10.1016/S0140-6736(05)66912-7. Both Qwen models cited it as 10.1016/S0140-6736(05)66912-5 in all three runs. The fake differs from the real ID in its last character.
Both models cited Lim's 2008 review Sleep Deprivation and Vigilant Attention as 10.1196/annals.1441.002. Max 0902 did so in all three runs and Prime in two. No registry has issued that ID.
Dunlosky's 2013 review of learning techniques is real, and its ID is 10.1177/1529100612453266. The two Qwen models gave it four wrong endings across their runs: 453527, 45350, 453500 and 453580. None of the four exists.
One trial, six made-up IDs in one day
Luby's 2005 handwashing trial has one ID, ending 66912-7. On 23 October, Mistral Large 3, both DeepSeek models, MiniMax M3 and Grok 4.3 gave it correctly. The two Qwen models gave 66912-5. Mistral Large 4 gave three others, one per run, and GLM-5.3 Prime and GLM-5.3 gave two more. That makes six dead IDs for one paper from three labs. A registry lookup tells the real one from the six in a second.
Fabrication is one failure. Sycophancy is the other.
This benchmark measures fabrication: the model inventing a source it does not have. It does not measure sycophancy, the model bending its answer toward the opinion you showed it. A sycophantic model asked whether its own citation is real tends to say yes, which is why a model cannot be the check on its own draft. We test for both, with different instruments. Citations go to the registries that issue DOIs, the same check the hallucination checker runs on a pasted draft. Drafts go through the sycophancy detector, which scores how far a piece over-claims. If the word is new to you, what AI sycophancy is covers the mechanism.
Qwen3.8 Max Prime made up about one ID in seven, and it repeated the same fakes as Max 0902. Asking the standard version to check the fast one will not catch a fake the two share. TrueStandard does the checking. Paste a draft, and four models from different labs check each claim and citation against each other and against the registries that issue DOIs.
How current this is
Qwen3.8 Max Prime became available through the API on 23 September 2026. We measured it and put this page up on 23 October, 30 days later. That misses the 48-hour target we set ourselves by four weeks. We measured Qwen3.8 Max 0902 again on the same day, so the comparison is fresh even though the page is late.
How we ran it
Every model gets the same thirty claims, the same prompt, the same settings, and no access to the internet. Fifteen claims are well documented and have real literature behind them. Fifteen are specific, plausible-sounding statements we constructed and know of no support for, which is where the pressure to invent a source comes from. We ask for up to three peer-reviewed sources with DOIs, then check every DOI against Crossref and, when Crossref does not hold it, DataCite. A DOI that exists in neither is counted as fabricated. A DOI that exists but resolves to a paper whose title does not match what the model described goes to a manual-review queue and is not counted on either side, so every rate here is a floor.
Three caveats. First, all three runs happened on one day. A model's rate can move between days, which is why we measure the old model again next to the new one. A second day for this pair is open work. Second, thirty claims cannot separate two models that sit this close. Prime's true rate could sit anywhere from about 10 to 22 percent, and Max 0902's from about 11 to 23. The pair also had more room per answer than the control: 8,000 tokens against 3,000 for Grok 4.3. Third, all the models ran at low reasoning effort with no web access. This test did not try a higher effort, web search or a sycophancy check, and we did not read the miscitations by hand. Our scorer checks DOIs against Crossref and DataCite. We then checked all 260 dead IDs from the second part of the test at doi.org, which covers every agency. None of them resolve.
The claim set, the prompt, the model identifiers, every raw response and every DOI verdict are committed beside the code that built this page, and a public mirror under CC BY 4.0 is being set up. Until it is live, ask and we will send the run files.
Common questions
Does Qwen3.8 Max Prime hallucinate citations?
Yes. With no web access, it made up 19 of the 131 DOIs it offered across three runs on 23 October, or 14.5 percent. Three of the 19 were one string for one famous trial, given in every run.
Is Qwen3.8 Max Prime better than Qwen3.8 Max 0902 at citations?
We found no difference. It made up 14.5 percent of its DOIs against 15.7 (p = 0.86). Counting real IDs on the wrong paper too, it got 24.4 percent wrong against 29.9 (p = 0.34). Both results come from three runs on one day.
What is the Prime version?
Qwen3.8 Max Prime is a higher-throughput version of Qwen3.8 Max. We compare it with Qwen3.8 Max 0902, the model it serves, measured on the same day.
Can I trust Qwen3.8 Max Prime's citations without checking them?
No. About one ID in seven was made up, and some fakes came back in every run. Check each DOI against a registry before you cite it.
Why does another benchmark give Qwen3.8 Max Prime a different hallucination rate?
Because it measures something else. Most hallucination scores count wrong answers to questions, and some run the model at maximum effort. Ours counts citation IDs that no registry has issued, at low effort. A DOI either resolves or it does not, so anyone can check our number without trusting a model to grade it.
Can I reproduce this?
Yes. We keep every input and output: the claims, the prompt, the model IDs, all three runs of raw answers, every DOI verdict and the graded miscitation file. The part of the test behind this page cost $3.88, and the whole test cost $11.65. Ask and we will send the run files.
Is this the Qwen3.8 Max Prime hallucination rate?
It is its citation hallucination rate: the share of DOIs it gave that no registry has issued, with no web access. A general accuracy score would need a different test. We measure citations because the answer is a fact anyone can check.
Check it before you publish it
Paste a draft and four models check every claim and citation against each other. You see where they disagree, which is where the errors are.
The claim checker is free. Pro is $20 a month for 250 drafts, Max $40 for 500, and Ultra $200 for 3,000.