“(In some cases o1 doesn’t even do better than GPT-4).”
Our reading, the sentence that was checked:
In some cases o1 does not perform better than GPT-4.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, The new AI scaling law shell game, paragraph 12.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-a-new-ai-scaling-law-shell-game-2dc649-a2
We recorded 12 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 2024-11-24. 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
- in some cases (qualifier, stated by the speaker)
From the speaker
- In some cases
- o1 doesn’t even do better than GPT-4
Our interpretation
- This is an existential comparison: at least one evaluated task or category where o1 ties or underperforms is sufficient.
- In the context of the cited research, GPT-4 includes the GPT-4o comparator, and o1 includes the contemporaneous o1-preview model where explicitly identified.
- Performance includes benchmark results and blinded human preference, as used in the cited evaluations.
- The target is the contemporaneous release period from the cited September research disclosure through the November post, not later model generations.
Counts as supported if: A contemporaneous evaluation identifies at least one task or category where the relevant o1 model does not outperform the GPT-4-family comparator.
Counts as contradicted if: The contemporaneous evidence establishes o1 superiority in all relevant evaluated cases and invalidates the alleged exceptions.
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
OpenAI reports o1 superiority in 54 of 57 MMLU subcategories, implying non-winning cases in the remainder. Its researchers also explicitly acknowledge tasks where the comparison model matches or surpasses o1. This supports Marcus's limited claim in context, with GPT-4 read as including GPT-4o. It does not establish that o1 is generally worse or that additional thinking time fails.
Limitations
- The supported comparison is principally against GPT-4o, not specifically the original GPT-4.
- The result applies to the contemporaneous models and evaluations.
- The evidence does not establish statistical significance for each individual non-winning category.
Evidence (2 items)
Supports the claim: OpenAI,
It also outperformed GPT‑4o on 54 out of 57 MMLU subcategories.
Why it matters here: The stated count implies that o1 did not outperform GPT-4o in the remaining three subcategories. It supports the limited 'some cases' claim under the contextual reading of GPT-4.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Supports the claim: OpenAI Forum,
And yeah, GPT-4.0 either matches or outshines the OpenAI O1 model in a few of these more abstract tasks as well.
Why it matters here: An OpenAI researcher explicitly acknowledges non-winning cases and directs listeners to the research post. The transcript's 'GPT-4.0' spelling should not be treated as proof that the comparator was the original GPT-4 rather than GPT-4o.
- Publisher
- OpenAI Forum
- 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, The new AI scaling law shell game
(In some cases o1 doesn’t even do better than GPT-4).
- Where
- paragraph 12
- Published
- 2024-11-24
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-a-new-ai-scaling-law-shell-game-2dc649-a2: 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-a-new-ai-scaling-law-shell-game-2dc649-a2.