5 articles

benchmarks

Index of editorial tags and filtered coverage.

Oct 9, 2026benchmarks

Humanity’s Sixth Sense: Scale AI’s new benchmark shows the best AI scores only 53.6% on visual intuition versus 93.1% for humans

Scale AI and Elorian introduced Humanity’s Sixth Sense, a 522-task benchmark with images and videos that measures everyday visual intuition: people score 93.1%, the best model, GPT-6-astra, scores 53.6% and the median of 25 models scores 30.9%.

$Artificial Analysis: GPT-6 Astra ties Claude Fable 5.1 at a fraction of the cost$

Sep 19, 2026Benchmarks

Artificial Analysis: GPT-6 Astra ties Claude Fable 5.1 at a fraction of the cost

September 2026 Artificial Analysis benchmarks place GPT-6 Astra and Claude Fable 5.1 tied near 53 on the Intelligence Index, with Astra markedly cheaper per task (~40–60% of the cost on price/token comparisons).

Sep 18, 2026MLPerf

MLPerf Inference v6.1: record participation with end-to-end RAG and edge agentic inference

MLCommons announced on 16 September 2026 MLPerf Inference v6.1 results with a record 30 organizations, new End-to-End RAG and Edge Agentic Inference tests, and up to 5.7× year-on-year gains in some scenarios—a different angle from the Vera Rubin NVL72 piece.

Sep 18, 2026Google

Gemini 3.8 Live and Live Extended Thinking: agentic voice with parallel reasoning

Google introduced on 15 September 2026 (updated 17) Gemini 3.8 Live and Live Extended Thinking: speech-to-speech models with Artificial Analysis 82.6, τ-Voice 68.6%, and τ-Voice-banking 35.1%, aimed at voice agents.

Sep 17, 2026RSIAgent

RSIAgent (arXiv:2609.15364): training-free multi-agent system with environment memory and claims vs GPT-6

RSIAgent (arXiv:2609.15364, 14 Sep 2026) proposes a training-free curriculum/actor/verifier with environment memory; on OSWorld-v2 and Agent's Last Exam the paper claims Kimi-K3 and GLM-5.3 beat some closed models including GPT-6—as a claim, not consensus.