Judging AI claims
Which AI has the highest IQ?
On the AI IQ rankings at aiiq.org, GPT-6.1 Sol and Opus 5.5 score highest at about 141, followed by Gemini 4 Argon and GPT-6 Astra at about 140. These are not human IQ scores, and test contamination makes comparisons contested.
The full answer
Top-tier AI models from OpenAI, Anthropic and Google lead current intelligence and IQ-style evaluations, with GPT-6.1 Sol and Opus 5.5 scoring around 141 on the AI IQ rankings at aiiq.org. According to that tracking data, the highest-scoring models are:
- GPT-6.1 Sol (OpenAI): about 141
- Opus 5.5 (Anthropic): about 141
- Gemini 4 Argon (Google): about 140
- GPT-6 Astra (OpenAI): about 140
- Gemini 4 Carbon (Google): about 139
Important context on AI "IQ".
- Not a human IQ. Standard human IQ tests (like Mensa) are designed for biological humans and rely on human cognitive architecture. When AI models "take" these tests, they are often evaluated on pattern recognition, logic and abstract reasoning benchmarks that mimic test questions.
- Contaminated training data. Many critics note that large language models may have met the test questions or similar logic puzzles in their web-scale training data, which makes direct comparisons to human scores controversial or misleading.
- Alternative benchmarks. Rather than traditional IQ, researchers evaluate AI with functional leaderboards such as FrontierMath, GPQA (graduate-level physics, chemistry and biology) and AIME (mathematical olympiad) to measure real problem-solving capability.
Read moreBenchmark contamination explainedReading benchmark claims criticallyIndependent AI leaderboardsGPT-6.1 Sol