Artificial Analysis: an index, ten tests, not a single score to trust
Artificial Analysis’s methodology page, read on 3 October 2026, describes Intelligence Index v4.3.2. Ten evaluations, including Terminal-Bench 4.0 and Humanity’s Last Exam. The index has limits. This is not a SecuFocus run.
Getting started
Original AI-generated illustration · SecuFocus
At a glance
Key points
An index summarizes. It does not say which model fits you. We did not rerun the ten tests.
What changes
The methodology page, read on 3 October 2026, says Index v4.3.2 combines ten evaluations: AA-Briefcase v1.1, GDPval-AA v2.1, AutomationBench-AA, Terminal-Bench 4.0, SciCode, AA-LCR v1.1, AA-Omniscience, Humanity’s Last Exam, GDP.pdf and CritPt.
On the homepage, Artificial Analysis says v4.3 replaces τ³-Banking with AutomationBench-AA and moves Terminal-Bench to 4.0. The homepage also announces a Cyber Index. That page does not carry the method detail read here.
What leaves
These tests send tasks to models, often through the provider’s API. The methodology passage we read does not list which conversations are kept. A public index does not say what your own use sends.
What we did not check
We did not rerun the ten tests, and we did not check the confidence interval they state. Artificial Analysis itself writes that the index has limits and does not apply to every use. A rank on that page is not a measurement made at your desk.
The choice
Read the index as a synthesis of ten tests, not as a truth. If the model you care about is not the same version, or not at the same effort setting, the rank does not travel. Keep the methodology page, not a screenshot of the podium.
Check and explore
Sources for this article
Numbers connect each reference to the passages that use it. Dates show when the documentation was consulted.
- Intelligence Benchmarking, Artificial Analysis ↗artificialanalysis.ai ·
This article draws on the sources above. The exercises are for you to try on your devices; SecuFocus does not present them as tests carried out by its editorial team. Interfaces and features can change. Method and corrections.
Cite this article
Keep this reference with the article when you save or share it.