AI NEWS PAGE

September 9, 2026 · Updated Sep 9, 3:59 PM PDT

BENCHMARKS:

DeepSWE — v1.1: GPT-6 Astra 74% · Gemini 3.8 Flash 74% · Claude Opus 5 74% · GPT-5.6 Sol 73% (Sep 3)Artificial AnalysisLM ArenaMETR Time HorizonsEpoch AI BenchmarksTerminal-Bench 2.0ARC-AGI