“Another way to think about this (see third graph above) is that there about 5 - 7 recent models are on a par with GPT-4, but none is clearly and decisively ahead.”
Our reading, the sentence that was checked:
About 5 to 7 recent models are on par with GPT-4, but none is clearly ahead.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Evidence that LLMs are reaching a point of diminishing returns - and what that might mean, paragraph 23.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-evidence-that-llms-are-reaching-a-5d187d-a1
We recorded 20 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-29. The source may have been edited after 2024-04-13. 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)
From the speaker
- about 5 - 7 recent models
- on a par with GPT-4
- none is clearly and decisively ahead
- see third graph above
Our interpretation
- The ordinary reading is that approximately five to seven recent models had broadly comparable capabilities to GPT-4, without a substantial overall leader.
- The count requires identifying whether separate snapshots count as separate models. Do not silently substitute a count of vendors.
- GPT-4 may mean the original model or the updated family; that distinction must remain explicit.
- For the ongoing comparison, use the latest pre-post release-and-evaluation window, March 1 through April 13, 2024. This reporting window is an evaluator choice.
Counts as supported if: A dated, identifiable model set contains approximately five to seven newer models with broadly comparable performance to the specified GPT-4 baseline, and no model has a robust substantial overall lead.
Counts as contradicted if: The dated model set materially fails the approximate count under a defensible parity definition, or a newer model has a robust substantial overall lead over the specified baseline.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Unresolved
The record was checked, and the evidence available by the cutoff does not settle it.
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
Contemporary records identify several GPT-4 contenders, but they do not verify the specific approximate count. The missing graph prevents identifying its members and baseline. Reports also distinguish original GPT-4 from later versions, which changes who qualifies as a peer or leader. Neither support nor contradiction of the full claim is established.
Limitations
- The earlier four-model list does not contradict a later approximate five-to-seven-model count.
- A ranking advantage is not automatically a decisive overall lead.
- Counting snapshots separately could materially change the result.
Evidence (2 items)
Context only: Simon Willison,
contender.
Why it matters here: The article names Gemini 1.5, Mistral Large, Claude 3 Opus and Inflection-2.5 as contenders. It supports the existence of a competitive group but is an earlier four-model assessment, not verification of Marcus's April count.
- Publisher
- Simon Willison
- Kind of source
- Secondary summary
- Describes events of
- Public from
Context only: The Decoder,
original GPT-4.
Why it matters here: This challenges treating all GPT-4 versions as one unchanged baseline. The report does not supply a parity band or enough uncertainty information to determine which models belong in the claimed count.
- Publisher
- The Decoder
- Kind of source
- News report
- Describes events of
- Public from
What would settle it
Recover the cited graph and its dated underlying data, identify the counted models and GPT-4 snapshot, and apply a consistent parity and substantial-lead assessment.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Evidence that LLMs are reaching a point of diminishing returns - and what that might mean
Another way to think about this (see third graph above) is that there about 5 - 7 recent models are on a par with GPT-4, but none is clearly and decisively ahead.
- Where
- paragraph 23
- Published
- 2024-04-13
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-evidence-that-llms-are-reaching-a-5d187d-a1: Current. Outcome: Unresolved. 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-evidence-that-llms-are-reaching-a-5d187d-a1.