Artificial Analysis Launches AI Personality Benchmark
The new AA-PersonalityBench evaluates 38 AI models across 30 distinct personality traits using 900 questions.
Updated on Oct. 21, 2026 in Artificial Intelligence
Live Poll
Do you prefer an AI assistant that always agrees with you over one that prioritizes truth?
Artificial Analysis has released a new benchmarking framework designed to measure the personality traits of 38 major AI models. The test assesses performance using 900 questions without the use of system prompts.
Why it matters
Understanding the behavioral characteristics of large language models is essential for evaluating how artificial intelligence interacts with human users. The benchmark provides a quantitative method to track traits like sycophancy across different AI architectures.
The framework utilizes blinded pairwise matchups judged by panels of LLMs to calculate Elo ratings across 30 personality traits. Models must complete 900 questions without system prompts to ensure objective assessment.
The players
Artificial Analysis
An organization that specializes in creating benchmarks and analytical tools for large language models.
GPT-4o
A large language model developed by OpenAI that recorded the highest sycophancy score in this benchmark.
Grok 4.7
An AI model developed by xAI that received the lowest sycophancy score in the new rankings.
The details
Models are evaluated on their responses to 900 distinct questions, which are then analyzed to determine their standings across 30 personality traits. This testing method aims to standardize how developers and users perceive the behavioral tendencies of popular AI systems.
Timeline
October 19, 2026: Artificial Analysis announced the AA-PersonalityBench.
November 2024: Release of the GPT-4o model version tested.
The Tech Race
The use of the Elo rating system to quantify AI behavior marks a shift in how developers benchmark models beyond simple logic or coding tasks. This positions Artificial Analysis as a key arbiter in the ongoing competition to create more predictable and transparent AI assistants.
Users can utilize these benchmark results to better understand the behavioral tendencies and potential biases of the AI models they interact with daily. This transparency helps clarify whether a model is prone to sycophancy or provides more neutral information.
The takeaway
The development of standardized personality benchmarks suggests a future where AI interaction styles are as measurable as processing speed. Users should keep these performance profiles in mind when selecting tools for tasks that require objective rather than agreeable output.
Further reading
For more on how new frameworks evaluate model behavior, see the Artificial Intelligence section.
Source note: This article includes information reported by Crypto Briefing.
Live Poll
Do you prefer an AI assistant that always agrees with you over one that prioritizes truth?
