A day after the busiest model launch window in AI history, the story shifted from announcements to verdicts — independent benchmarks on Grok 4.5 landed, an AI-driven chipmaker set a Wall Street record, and state regulators kept moving.
Grok 4.5 gets its first independent report card
Twenty-four hours after launch, Artificial Analysis ranked xAI’s Grok 4.5 fourth on its Intelligence Index, behind Claude Fable 5, GPT-5.5, and Claude Opus 4.8 — while awarding it the best agentic tool-use score of any model on the board. The headline caveat: the model’s hallucination rate jumped from 25% to 54% versus Grok 4.3, meaning it knows more but is more confidently wrong when it errs. First-day reviews suggest it is a strong fit for tool-calling agent workflows, and a risky one where output accuracy is the metric.
Snorkel: Grok 4.5 leads on real professional work
Snorkel AI’s independent evaluation on roughly 2,000 expert-authored GDPval+ tasks put Grok 4.5 at a 29% mean pass rate, ahead of GPT-5.5 (22%) and Claude Opus 4.8 (21%). The lead was concentrated in domains requiring professional judgment — legal work, education, and healthcare documentation. Snorkel’s analysis is the most credible early signal on how the new model handles actual workplace deliverables rather than synthetic puzzles.
SK Hynix rides the AI boom to a record US listing
Memory chipmaker SK Hynix — a key supplier for AI data centers — closed its first Nasdaq session up 13% at $168.01 after raising $26.5 billion, the largest US share sale ever by a foreign company. CNBC reports the proceeds will fund aggressive factory and equipment expansion, with the chairman telling investors that AI-driven memory demand “is enormous.”
Illinois signs a landmark state AI safety law
Governor JB Pritzker signed Illinois’s AI Safety Measures Act into law this week, adding another major state to the growing patchwork of US AI regulation. The Transparency Coalition’s legislative update notes that companion-chatbot rules and training-data transparency remain the most active areas of state lawmaking at 2026’s halfway mark.
What to watch: Google’s Gemini 3.5 Pro is now six weeks past its GA target with no confirmed date — every day of delay raises the stakes for its eventual debut.