“A new study of the latest system, GPT 4.5 showed that by on one standardized measure over 1/3 of the outputs are still hallucinations.”
Our reading, the sentence that was checked:
On one standardized measure, over a third of GPT-4.5's outputs are hallucinations.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Decoding (and debunking) Hard Fork’s Kevin Roose, paragraph 16.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-decoding-and-debunking-hard-forks-a7e70f-a1
We recorded 16 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2025-03-02. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Observation
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No horizon
- Deadline state on 2026-10-29
- No horizon (the release’s reference date)
Conditions
- on one standardized measure (qualifier, stated by the speaker)
From the speaker
- A new study
- GPT 4.5
- on one standardized measure
- over 1/3
- outputs are still hallucinations
Our interpretation
- The claim concerns a contemporaneous reported benchmark result, not all GPT-4.5 outputs in ordinary use.
- SimpleQA is the identified measure corresponding to the reported result.
- The target period is February 2025, the launch-evaluation month immediately preceding the post.
- Over one third requires a hallucination rate strictly greater than 33⅓%.
Counts as supported if: The contemporaneous GPT-4.5 evaluation reports a hallucination rate greater than one third on at least one standardized benchmark.
Counts as contradicted if: The identified result is at or below one third, concerns another model, or measures something other than the claimed hallucination rate.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Supported
The dated evidence matches what the claim committed to, under the criteria frozen before grading.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
The reported SimpleQA hallucination rate was 37.1%, which exceeds one third. OpenAI's launch page confirms the benchmark and its factual-question scope. Marcus expressly restricted the statement to one standardized measure, so it should not be read as a claim about all GPT-4.5 responses.
Limitations
- The exact percentage is supported by Computerworld's report of OpenAI's evaluation, not a readable primary numeric table.
- This is a benchmark-specific result and does not establish an ordinary-use hallucination rate.
- The cited news report was published later but describes the February model evaluation.
Evidence (2 items)
Supports the claim: Computerworld,
37.1% hallucination rate on SimpleQA.
Why it matters here: The report identifies a 37.1% hallucination rate for the February GPT-4.5 model on SimpleQA, exceeding one third. This is secondary reporting of OpenAI's result because the primary page's numeric chart values were absent from readable text.
- Publisher
- Computerworld
- Kind of source
- News report
- Describes events of
- Public from
Context only: OpenAI,
challenging knowledge questions.
Why it matters here: The launch announcement confirms the relevant standardized evaluation and its factual-question scope. The extracted text does not expose the numeric bars, so this item supplies context rather than the threshold evidence.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Decoding (and debunking) Hard Fork’s Kevin Roose
A new study of the latest system, GPT 4.5 showed that by on one standardized measure over 1/3 of the outputs are still hallucinations.
- Where
- paragraph 16
- Published
- 2025-03-02
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-decoding-and-debunking-hard-forks-a7e70f-a1: Current. Outcome: Supported. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
About assessment marcus-decoding-and-debunking-hard-forks-a7e70f-a1.