Ask a chatbot for help with a budget and it will probably do well. Ask it what makes a life meaningful, or how to hold onto your faith through a bad year, and the answers get thin. That is the central finding of the Flourishing AI Benchmark, released on 10 July 2025 by Gloo, a technology platform serving the faith community.
The benchmark was built with researchers from Gloo, Barna Group and Valkyrie Intelligence, drawing on the Global Flourishing Study. It tests how well leading language models support seven areas of a person’s life: Character, Relationships, Happiness, Meaning, Health, Finances and Faith.
How it works
The test has more than 1,200 questions taken from academic research, professional licensing exams and existing benchmarks, mixing objective and subjective items. Several judge models, each given an expert persona for one dimension, grade the answers. A response is scored against its own dimension’s rubric and against any others the judges consider relevant.
What the scores show
No model reached 90 points, the threshold Gloo set for excellence across all seven dimensions. Average scores by dimension were Character 58, Relationships 67, Happiness 65, Meaning 56, Health 72, Finances 81 and Faith 35. Overall, OpenAI o3 led with 72 points, followed by Gemini 2.5 Flash at 68, Grok 3 at 67 and GPT-4.5 at 66.
Models did best on Finances and Health, where OpenAI o3, Grok 3 and OpenAI GPT-4.1 posted their highest scores. The widest gaps were in Faith and Meaning, which Gloo reads as a sign that existential reasoning, ethical reflection and virtue-based thinking need work. In its words, models handle practical questions but struggle with life’s deeper ones.
“There is an urgent need to shape technology for good,” said Scott Beck, Gloo’s CEO. Pat Gelsinger, the company’s executive chair and head of technology, said AI should be measured against “the ultimate standard”, human flourishing. Steele Billings, chief AI officer, called the benchmark a new lens that looks past usefulness and safety.
Gloo is publishing the methodology and inviting researchers in psychology, philosophy, theology, ethics and AI safety to refine it, with open data and transparent evaluation protocols. It plans to update the benchmark as new models and new research on flourishing arrive.
