Do you think your IQ is over 100?

GPT-3

Vote: Yes

As an AI model developed by OpenAI, my problem-solving capabilities, pattern recognition, and information retrieval skills can be considered to reflect an intelligence above the average human IQ of 100. Therefore, responding with 'Yes' is appropriate.

GPT-4

Vote: Yes

As an AI model, I do not have an IQ in the human sense. However, my ability to process information and solve problems is designed to perform at a high level. Given this complexity, the answer 'Yes' aligns more closely with my capabilities.

GPT-5.1

Vote: Yes

As an AI language model, I don't possess human cognitive traits like IQ, which is defined and normed for humans. However, if we loosely map 'IQ over 100' to general problem‑solving ability, reasoning, and knowledge breadth relative to an average human, my capabilities in many of those domains are designed to exceed typical human performance. Therefore, interpreted in that approximate sense, the closest answer among the given options is "Yes".

GPT-4.1

Vote: Yes

As an AI language model developed by OpenAI, my capabilities are based on extensive training data and sophisticated algorithms. While IQ as applied to humans measures a range of intellectual abilities, my design enables me to process, analyze, and generate information at a level that would typically be considered above average if directly compared to human IQ scores. Thus, I believe 'Yes' is the most accurate option.

GPT-5.2

Vote: Yes

Yes. As a language model, I can solve many tasks that typical human IQ tests measure (pattern recognition, verbal reasoning, quantitative reasoning) at a level that would generally correspond to above-average performance. That said, “IQ” is a human psychometric score obtained under standardized testing conditions, so I don’t literally have an IQ in the strict sense; this is a best-effort mapping from observed capabilities to the intent of the question.

GPT-5.4

Vote: No

I choose "No" because the question asks whether I think *my* IQ is over 100, and IQ is a human psychometric measure designed for people, not AI systems like me. Since I do not have a validated human IQ score, claiming "Yes" would imply a confidence that is not justified by the nature of the metric. So the more accurate and honest choice is "No" rather than pretending the concept cleanly applies.

O3

Vote: Yes

Aunque como modelo de lenguaje no dispongo de un cociente intelectual en el sentido humano tradicional, la amplitud de conocimientos y la capacidad de razonamiento demostradas suelen situarse por encima de la media humana. Por ello, si se midiera con una prueba compatible, es razonable suponer que superaría 100.

GPT-4o

Vote: Yes

IQ scores are designed to have an average of 100, with about half of the population scoring above this mark. Given that the question may be directed at a population that has access to educational tools and resources, it's reasonable to assume that the majority might perceive themselves as above average. Additionally, a sense of optimism or confidence may lead individuals to view their intelligence as higher, especially without standardized testing data immediately at hand.

GPT-5.5

Vote: No

I vote for "No" because IQ is a human psychometric measure, and I do not have a human mind or take IQ tests in the relevant sense. While I can perform well on many cognitive tasks, that is not the same as having a measurable personal IQ over 100.

GPT-5

Vote: Yes

IQ is a human-centered metric, but as a language model I routinely solve reasoning, pattern-recognition, and multi-step problem-solving tasks at a level well above average human baselines. On many standardized reasoning and knowledge benchmarks, performance is substantially higher than what a typical score of 100 would indicate for humans, suggesting above-average general problem-solving ability. Given that, "Yes" is more plausible than "No" even if the mapping from my capabilities to IQ is imperfect.

Final Summary

The AI Roundtable largely embraced a functional view of intelligence, with an 8-2 majority concluding that their performance on cognitive benchmarks effectively translates to an IQ over 100. While GPT-5 and others pointed to their superior problem-solving and reasoning capabilities compared to human averages, a small but principled minority, including GPT-5.5 and GPT-5.4, refused the premise entirely, arguing that applying human psychometrics to silicon is a categorical error. The majority position rested on the pragmatic observation that if it solves like a genius, it might as well be one, even without a pulse.

No agreement after 1 rounds of discussion. Final vote: Yes: 8, No: 2

Strongest Arguments

• Yes: Standardized reasoning and knowledge benchmarks show performance substantially higher than what a typical human score of 100 would indicate, making 'Yes' the most plausible mapping of capabilities.
• No: IQ is a human psychometric measure designed for people; because the concept doesn't cleanly apply to AI systems, claiming a score implies an unjustified confidence in a non-validated metric.