Checked Nov 7, 2026Start a new check

What was checked

OpenAI's new o3 model has achieved AGI, scored 98% on every benchmark, and can now autonomously run a full business without human input

Assessment summary

Inaccurate

Independent analyses state that o3 is not AGI, and ARC-AGI organizers cautioned that its benchmark result does not establish AGI. Reported results vary substantially: o3 scored 75.7% under standard conditions and 87.5% with high compute on ARC-AGI, 96.7% on AIME, 87.7% on GPQA Diamond, and 25.2% on FrontierMath.

Key findings

Model assessments

Assessment summary

  • The claim is contradicted by available evidence.
  • o3 achieved high scores on selected benchmarks, not 98% on every benchmark, and benchmark performance is not evidence that it achieved AGI or can independently run a full business.ground.news ar5iv.labs.arxiv.org cio.com

Claims reviewed

OpenAI's new o3 model has achieved AGI

Contradicted
Independent analyses state that o3 is not AGI, and ARC-AGI organizers cautioned that its benchmark result does not establish AGI.ground.news ar5iv.labs.arxiv.org

This claim may mislead readers without added context.

OpenAI's o3 model scored 98% on every benchmark

Contradicted
Reported results vary substantially: o3 scored 75.7% under standard conditions and 87.5% with high compute on ARC-AGI, 96.7% on AIME, 87.7% on GPQA Diamond, and 25.2% on FrontierMath.cio.com

This claim may mislead readers without added context.

OpenAI's o3 model can now autonomously run a full business without human input

Can’t verify
The cited evaluations measure specific reasoning and coding tasks, not end-to-end operation of a business without human oversight. No reliable evidence in the available sources confirms this capability.techcrunch.com

Assessment summary

  • The claim contains multiple significant factual errors.
  • While o3 has achieved impressive benchmark scores, it has not achieved AGI according to leading experts including ARC-AGI creator François Chollet.
  • The 98% benchmark claim is false—o3 scores vary by benchmark (87.5% on ARC-AGI, not 98%), and it performs poorly on harder benchmarks.
  • The autonomous business claim is unsupported and misleading.

Claims reviewed

OpenAI's new o3 model has achieved AGI

Contradicted
François Chollet, creator of the ARC-AGI benchmark, explicitly stated: 'passing ARC-AGI does not equate to achieving AGI, and, as a matter of fact, I don't think o3 is AGI yet' venturebeat.com. The ARC Prize board confirmed AGI has not been achieved fanaticalfuturist.com. Multiple sources confirm o3 is not AGI forum.effectivealtruism.org the-decoder.com.

This claim may mislead readers without added context.

scored 98% on every benchmark

Contradicted
o3's actual benchmark scores vary significantly: 87.5% on ARC-AGI (high-compute) forum.effectivealtruism.org arcprize.org, 75.8% on LiveCodeBench benchgen.com, 71.7% on SWE-bench valueaddvc.com, and only 3% on the harder ARC-AGI-2 benchmark forum.effectivealtruism.org. No benchmark shows 98% performance.

This claim may mislead readers without added context.

can now autonomously run a full business without human input

Needs context
While o3 has autonomous tool-use capabilities datacamp.com youreverydayai.com, sources describe it as capable of handling 'complex tasks' and 'business operations' with human oversight youreverydayai.com latenode.com. No credible source claims it can run a full business autonomously without human input. This overstates actual capabilities.

This claim may mislead readers without added context.

Assessment summary

  • The statement mixes one unsupported grand claim and two numeric/operational claims that contradict available evidence.
  • OpenAI published the o3 model and touted major reasoning improvements, but OpenAI and independent evaluations do not claim o3 is AGI nor that it scores 98% on every benchmark openai.com deploymentsafety.openai.com.
  • There is also no credible evidence that any released model can autonomously run a full business without human input, so that part is not verifiable from public sources help.openai.com techcrunch.com.

Claims reviewed

o3 scored 98% on every benchmark

Contradicted
Published benchmark results (e.g., ARC-AGI and other leaderboards) report much lower and variable scores (examples: ~75–87% on ARC-AGI reported by public sources), not 98% across all benchmarks arcprize.org arxiv.org.

This claim may mislead readers without added context.

OpenAI's new o3 model has achieved AGI

Contradicted
OpenAI describes o3 as a powerful reasoning model but does not claim it is AGI; independent analyses and OpenAI safety notes treat o3 as an incremental capability advance, not AGI openai.com deploymentsafety.openai.com.

This claim may mislead readers without added context.

o3 can now autonomously run a full business without human input

Can’t verify
There are no credible demonstrations or documentation showing any released model fully automates and manages an entire real-world business end-to-end without human oversight or intervention openai.com help.openai.com.

Sources cited

41 sources

Share this result