What changes were made in Intelligence Index v4.2?

Artificial Analysis added two new benchmarks: AA-Briefcase for knowledge work and GDP.pdf for PDF reading. It removed GPQA Diamond because model scores were saturated, and doubled the share of private data to 40%.

Which AI models lead the updated ranking?

Anthropic's Claude Fable 5.1 is the top model, followed by OpenAI's GPT-6 Astra, which gained 4 points over GPT-5.6 Sol.

Large Language ModelsAI Safety & AlignmentITmedia AI＋Sep 7, 2026Sep 7, 2026, 13:00 JST

Claude Fable 5.1 tops new v4.2 Intelligence Index

Send on LINE

3 Key Points

What happened
Artificial Analysis released Intelligence Index v4.2 on September 4, adding AA-Briefcase and GDP.pdf tests while dropping GPQA Diamond, which had saturated scores.
Why it matters
Private data now makes up 40% of the index, double v4.1's share, aiming to prevent AI developers from gaming benchmarks; v5 will raise it further.
What to watch
GPT-6 Astra rose 4 points from GPT-5.6 Sol, and showed higher output token efficiency than other frontier models.

WHO IT HITSEnterprise AI buyers and technical evaluators who rely on benchmark rankings to choose between large language models will now see scores less affected by AI developers' test-specific optimization.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Get news like this every morning, free

Context & Analysis

Artificial Analysis's v4.2 update arrives as AI developers commonly train models against public benchmarks, pushing scores higher without reflecting real-world usefulness. By introducing AA-Briefcase and GDP.pdf, and removing GPQA Diamond, the firm aims to capture abilities that were not being measured well.

The heavier weight on private test data represents a structural shift in how this index gauges model quality. The publisher states the purpose is to prevent benchmark gaming, and it plans to increase the private-data share even more in v5. This suggests ongoing cat-and-mouse dynamics between evaluators and developers.

For buyers selecting AI models for enterprise tasks, the stakes are in whether the new tests genuinely track practical performance. The outcome hinges on how well AA-Briefcase and GDP.pdf resist optimization compared to older benchmarks, possibly determining whether the top rank reflects true capability or just better test preparation.

FAQ

What changes were made in Intelligence Index v4.2?: Artificial Analysis added two new benchmarks: AA-Briefcase for knowledge work and GDP.pdf for PDF reading. It removed GPQA Diamond because model scores were saturated, and doubled the share of private data to 40%.
Which AI models lead the updated ranking?: Anthropic's Claude Fable 5.1 is the top model, followed by OpenAI's GPT-6 Astra, which gained 4 points over GPT-5.6 Sol.

ITmedia AI＋Read Original Article

Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.