“The best results were truly impressive, such as major advances on Francois Chollet’s ARC task and on an important mathematical benchmark FrontierMath.”
Our reading, the sentence that was checked:
o3's best demo results showed truly impressive major advances on ARC and FrontierMath.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, o3, AGI, the art of the demo, and what you can expect in 2025, paragraph 3.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-o3-agi-the-art-of-the-demo-and-what-9a94fb-a2
We recorded 15 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 2024-12-21. 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
- best results (qualifier, stated by the speaker)
From the speaker
- The best results were truly impressive
- major advances on Francois Chollet's ARC task
- and on an important mathematical benchmark FrontierMath
Our interpretation
- 'Truly impressive' is a judgment and is not scored on its own. The checkable part is whether o3's reported scores on ARC-AGI and FrontierMath at the 20 December 2024 reveal were large jumps over the previous best results.
- 'Major advance' means a jump well beyond small, step-by-step gains over earlier models.
Counts as supported if: Scores reported at the reveal show o3 far above earlier models on both ARC-AGI and FrontierMath.
Counts as contradicted if: Scores reported at the reveal show only small gains over the previous best, or no gain, on either benchmark.
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
On 20 December 2024, ARC Prize reported that o3 scored 75.7% on the ARC-AGI-1 Semi-Private set within its $10k public leaderboard compute limit. A high-compute configuration scored 87.5% (e1). ARC Prize said ARC-AGI-1 went from 0% with GPT-3 in 2020 to 5% with GPT-4o in 2024, and it called o3's result a breakthrough (e1, e2). TechCrunch reported the same day that o3 solved 25.2% of FrontierMath problems and that no other model exceeded 2%. That figure came from OpenAI's internal evaluations (e3). Both reported results were large jumps over earlier models, so the checkable part of the claim holds. On 18 April 2025, Epoch AI's results put o3 closer to 10% on FrontierMath (e4). TechRepublic said the December test used a prerelease version of o3 and that the problem set had changed since then (e4). That later result concerns a different setup and does not change what was reported at the reveal.
Limitations
- 'Truly impressive' is a judgment and was not scored.
- The December FrontierMath figure was OpenAI's own internal evaluation. The packet has no primary Epoch AI confirmation from that time.
- ARC Prize states that the o3 it tested was trained on the ARC-AGI-1 Public Training set. The evidence does not show how this affected the score.
- The evidence covers results reported at the reveal. It does not show how the released o3 performs across tasks in general.
Evidence (4 items)
Supports the claim: ARC Prize Foundation,
has scored a breakthrough 75.7% on the Semi-Private Evaluation set at our stated public leaderboard $10k compute limit. A high-compute (172x) o3 configuration scored 87.5%.
Why it matters here: The benchmark owner verified o3's ARC-AGI-1 scores of 75.7% and 87.5% at the reveal.
- Where
- opening section
- Publisher
- ARC Prize Foundation
- Kind of source
- Benchmark result
- Describes events of
- Public from
Supports the claim: ARC Prize Foundation,
For context, ARC-AGI-1 took 4 years to go from 0% with GPT-3 in 2020 to 5% in 2024 with GPT-4o.
Why it matters here: Gives the earlier baseline. The o3 result was a step change over the GPT series.
- Publisher
- ARC Prize Foundation
- Kind of source
- Benchmark result
- Describes events of
- Public from
Supports the claim: TechCrunch,
o3 sets a new record on EpochAI’s Frontier Math benchmark, solving 25.2% of problems; no other model exceeds 2%.
Why it matters here: The FrontierMath score reported at the reveal was more than ten times the previous best.
- Publisher
- TechCrunch
- Kind of source
- News report
- Describes events of
- Public from
Context only: TechRepublic,
On April 18, Epoch AI released test results showing OpenAI o3 scored closer to 10%.
Why it matters here: An independent retest of the released o3 found a lower FrontierMath score. It used a different model version and problem set, so it does not change what the demo showed.
- Publisher
- TechRepublic
- Kind of source
- News report
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, o3, AGI, the art of the demo, and what you can expect in 2025
The best results were truly impressive, such as major advances on Francois Chollet’s ARC task and on an important mathematical benchmark FrontierMath.
- Where
- paragraph 3
- Published
- 2024-12-21
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-o3-agi-the-art-of-the-demo-and-what-9a94fb-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-o3-agi-the-art-of-the-demo-and-what-9a94fb-a2.