Photo: Steve A Johnson / Pexels
State of AI report returns for its ninth year with Anthropic atop the benchmarks
Nathan Benaich's annual review frames the frontier as a three-lab race and leans on Artificial Analysis data to keep score
The State of AI Report is back for its ninth edition. This year’s version frames the frontier of artificial intelligence as a three-horse race between Anthropic, OpenAI, and Google.
Leading that pack, at least on the scoreboard the report chose to use, is Anthropic.
What the 2026 report covers
The report was published on October 8, 2026, by Nathan Benaich of Air Street Capital. It is an independent annual review that has run every year since its debut in 2018.
Its scope is broad. The report tracks developments across AI research, industry, geopolitics, safety, and policy.
To rank the leading models, this year’s edition pulls benchmark data from Artificial Analysis and its Intelligence Index. The report draws on versions v4.1 through v4.3 of that index, which focus on agentic workloads.
On those measures, the report places Anthropic in the lead on the Intelligence Index. OpenAI and Google round out what the report characterizes as the competitive frontier.
The four big themes
Beyond the leaderboard, the report highlights four themes that its authors see shaping the field.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
AI helping build AI. The first is the growing role AI plays in its own development. Models are increasingly part of the toolkit used to create the next generation of models.
Physical AI and robotics. The second theme is progress in physical AI and robotics.
An inference surge. The third theme is a sharp rise in inference workloads. Training is the expensive process of teaching a model. Inference is the ongoing work of actually running it every time someone asks it to do something.
Rising cyber threats. The fourth theme is heightened cyber risk.
Grading last year’s homework
One of the report’s recurring features is an honest look at its own track record. Each year it scores the predictions it made the year before.
This edition’s updated scorecard shows two correct hits, five partial hits, and three misses. Across those ten graded calls, outright wins were the smallest category.
The report also offers forward-looking views on geopolitics and safety, setting up next year’s round of grading.
Why a nine-year-old report still matters
When the State of AI Report launched in 2018, the field looked very different. Over nine editions it has become a recurring checkpoint for researchers, founders, investors, and policymakers trying to make sense of a fast-moving industry.