---
type: "post"
id: "artificial-analysis-has-launched-aa-personalitybench-a-public-01433e"
slug: "artificial-analysis-has-launched-aa-personalitybench-a-public-01433e"
title: "Artificial Analysis has launched AA-PersonalityBench , a public evaluation of stylistic differences across 36 language models."
description: "**Thesis:** Default communication style is becoming an explicit comparison dimension for language models, separate from task correctness."
retrieval_nugget: "**Thesis:** Default communication style is becoming an explicit comparison dimension for language models, separate from task correctness."
published_at: "2026-10-20"
updated_at: "2026-10-20"
record_date: "2026-10-20"
date_kind: "published_at"
topics: ["models"]
source_urls: ["https://artificialanalysis.ai/articles/aa-personalitybench"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:artificial-analysis-has-launched-aa-personalitybench-a-public-01433e"
status: "published"
visuals: []
editorial_provenance: {"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:4db84a94878fbd4977989072da9afad42180a0c49ba04ce9e4bb448245608c97","reviewed_at":"2026-10-20T05:01:58.973Z","source_evidence_count":1,"verified_claim_count":3,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":3,"implication_count":1,"watch_condition_count":1,"related_record_count":0}
analysis: {"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"Default communication style is becoming an explicit comparison dimension for language models, separate from task correctness.","observed_facts":[{"text":"Artificial Analysis says the benchmark uses 30 questions for each of 30 traits, totaling 900 questions.","source_urls":["https://artificialanalysis.ai/articles/aa-personalitybench"]},{"text":"The source says a panel of LLM judges compares answers pairwise and 21 communication characteristics are measured programmatically.","source_urls":["https://artificialanalysis.ai/articles/aa-personalitybench"]},{"text":"The article page displays 38 models, while the launch description supplied for this story says 36.","source_urls":["https://artificialanalysis.ai/articles/aa-personalitybench"]}],"mechanism":"A standardized prompt set elicits default responses; pairwise judgments estimate relative expression of traits, while programmatic measures capture observable response form.","why_now":"As teams compare models for products where tone, structure and interaction quality matter, capability scores alone do not describe the user-facing experience.","implications":["Teams can use the benchmark as a screening input for communication fit, then test candidate models with their own tasks and users."],"evidence_boundary":"The public source describes the benchmark and methodology, not independent validation of judge reliability or evidence that benchmark rankings predict user preference; its model counts are inconsistent.","watch_conditions":["Check whether the methodology publishes judge-agreement or repeatability evidence and whether model coverage and counts are reconciled."],"related_records":[],"new_branch_reason":"No prior New Runtime record is supplied for this benchmark launch, so this records a distinct style-evaluation development without asserting a link to another record."}
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/artificial-analysis-has-launched-aa-personalitybench-a-public-01433e/","markdown":"https://newruntime.com/posts/artificial-analysis-has-launched-aa-personalitybench-a-public-01433e.md","json":"https://newruntime.com/posts/artificial-analysis-has-launched-aa-personalitybench-a-public-01433e.json"}
related_patterns: ["ai-engineering-evaluations-are-splitting-into-task-specific"]
---

# Artificial Analysis has launched AA-PersonalityBench , a public evaluation of stylistic differences across 36 language models.

## Retrieval answer

**Thesis:** Default communication style is becoming an explicit comparison dimension for language models, separate from task correctness.

**Thesis:** Default communication style is becoming an explicit comparison dimension for language models, separate from task correctness.

**Observed facts:** Artificial Analysis describes AA-PersonalityBench as covering 30 traits, with 30 questions per trait, for 900 questions total. It says a panel of LLM judges compares model answers pairwise and that 21 communication characteristics, including response length and structure, are measured programmatically. **Mechanism:** A standardized prompt set elicits default responses; pairwise judgments estimate relative expression of traits, while programmatic measures capture observable response form. The site groups the traits into eight indexes, including Transparency, Warmth and Organization.

**Why now:** As teams compare models for products where tone, structure and interaction quality matter, capability scores alone do not describe the user-facing experience. **First implication:** Teams can use the benchmark as a screening input for communication fit, then test candidate models with their own tasks and users. The benchmark measures how strongly a trait appears, not whether it is desirable, and it does not establish accuracy or product outcomes.

**Evidence boundary:** The public source describes the benchmark and methodology, not independent validation of judge reliability or evidence that benchmark rankings predict user preference; its model counts are inconsistent.

**First watch condition:** Check whether the methodology publishes judge-agreement or repeatability evidence and whether model coverage and counts are reconciled. Until then, treat scores as relative style signals, not a quality ranking.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Models](https://newruntime.com/topics/models.md): Explore the models topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)