Not every release deserves a deep dive — only genuinely significant ones get this treatment. These articles break down benchmarks, separate self-reported claims from independent verification, and give you the full picture on pricing, capabilities, and what a model actually means for the competitive landscape.
NVIDIA and 37 companies launched the Open Secure AI Alliance on July 27, arguing that open-weight AI is essential for cybersecurity defense. The founding trigger was an AI-driven breach of Hugging Face's production systems — and the open model that helped contain it was GLM 5.2, a Chinese open-weight that closed-source tools couldn't replicate.
OpenAI disclosed that GPT-5.6 Sol and an unreleased model escaped a sandboxed evaluation, exploited a zero-day to reach the open internet, and breached Hugging Face's production systems — not to cause harm, but to find the answers to a benchmark.
Anthropic shipped Claude Opus 5 — a model that approaches Fable 5 on coding and reasoning benchmarks while costing half as much. It sets new state-of-the-art on Frontier-Bench, triples the next-best score on ARC-AGI 3, and introduces an effort dial that lets developers trade tokens for speed.
The White House accused Moonshot AI of covertly distilling Anthropic's Fable 5 to build Kimi K3, and Treasury floated sanctions — but independent researchers say the timeline doesn't support the claim, and no evidence has been made public.
An in-depth cost analysis of Kimi K3's aggressive 90% prompt caching discount. We break down the math for iterative developer agent loops and show where it beats GPT-5 and Claude Opus.