Artificial Analysis Agentic Index cited for a new Qwen 'Max' top ranking — unverified from source

A Hacker News post points to Artificial Analysis, an independent AI benchmarking site, claiming a Qwen ‘Max’ model has taken the top spot on its Agentic Index. The linked page, however, is the site’s index of leaderboards rather than a results table: it lists their evaluation suites — the Intelligence Index, Agentic Index, Coding Agent Index, AA-Briefcase (long-horizon knowledge work), AA-Omniscience (knowledge and hallucination), GDPval-AA (economically valuable real-world tasks), and the Openness Index — alongside cost, output-token, and latency comparisons. None of the captured content includes the actual ranking, scores, or any reference to Qwen.

Artificial Analysis’s Agentic Index is meant to score models on realistic, tool-using, multi-step work — agentic tool use, coding and terminal tasks, SaaS and business-operations workflows, and long-horizon tasks — normalized on an Elo-derived scale. A shift at the top of that board would be notable because agentic capability, not raw benchmark ‘intelligence,’ is what matters most for autonomous coding and workflow agents.

That said, the specific claim cannot be confirmed from the provided material, and the cited model string (‘Qwen3.8 Max’) does not correspond to Alibaba’s known Qwen naming. Because these leaderboards update continuously, any top-ranking claim should be checked live against the page before publication.