“But still part of the pack, not a giant leap forward (e.g. Grok 4 beats it on ARC-AGI-2 results)”
Our reading, the sentence that was checked:
GPT-5 is still part of the pack rather than a giant leap forward; for example, Grok 4 beats it on ARC-AGI-2.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, GPT-5 hot take, paragraph 5.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-gpt-5-hot-take-6bff2b-a1
We recorded 8 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 2025-08-07. 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
- still part of the pack
- not a giant leap forward
- Grok 4 beats it on ARC-AGI-2 results
Our interpretation
- Read 'part of the pack' as remaining within the contemporary frontier-model performance range, even if GPT-5 leads some evaluations.
- Read 'not a giant leap' relative to models available at launch, as indicated by the Grok 4 comparison, rather than relative only to the original GPT-4.
- Assess the ARC-AGI-2 example by reported accuracy for the launch configurations. This does not imply equal inference cost or overall model superiority.
- Use multiple benchmark perspectives without inventing a numerical threshold for 'giant'.
Counts as supported if: Grok 4 exceeds GPT-5 on contemporaneous ARC-AGI-2 accuracy, and broader launch evaluations show gains without a clear general separation from the existing frontier.
Counts as contradicted if: The cited ARC-AGI-2 comparison is reversed under the relevant launch settings, or broad comparable evaluations establish a clear general separation from contemporary frontier models.
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
Supported under the contemporary-frontier reading. The ARC Prize founder reported Grok 4 at 15.9% and GPT-5 at 9.9% on ARC-AGI-2. Artificial Analysis independently found a new overall high for GPT-5 but a smaller advance than earlier major transitions. This supports continued close competition, not an absence of progress. Epoch's comparison with GPT-4 shows why the conclusion changes if 'giant leap' instead means improvement across entire GPT generations.
Limitations
- The broader conclusion depends on the stated comparison baseline.
- GPT-5 led some evaluations; Marcus's wording should not be read as denying that.
- ARC-AGI-2 accuracy alone does not establish overall superiority or cost efficiency.
Evidence (3 items)
Supports the claim: François Chollet,
15.9% for Grok 4 vs 9.9% for GPT-5.
Why it matters here: The benchmark founder directly confirms Marcus's numerical example. This establishes accuracy ordering on this benchmark, not general model superiority.
- Publisher
- François Chollet
- Kind of source
- Benchmark result
- Describes events of
- Public from
Supports the claim: Artificial Analysis,
Medium (67) is close to o3
Why it matters here: The same report places High at a new index peak of 68 but describes its increase over o3 as smaller than earlier major model transitions. This supports a limited advance over the contemporary frontier while acknowledging leadership on the index.
- Publisher
- Artificial Analysis
- Kind of source
- Independent measurement
- Describes events of
- Public from
Context only: Epoch AI,
both large advances from the previous generation
Why it matters here: This later analysis finds large gains relative to GPT-4 and explains that intermediate releases spread those gains over time. It challenges a broader reading that denies substantial generational progress, but uses a different comparison baseline.
- Where
- Opening paragraph
- Publisher
- Epoch AI
- Kind of source
- Independent measurement
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, GPT-5 hot take
But still part of the pack, not a giant leap forward (e.g. Grok 4 beats it on ARC-AGI-2 results)
- Where
- paragraph 5
- Published
- 2025-08-07
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-gpt-5-hot-take-6bff2b-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.