Fresh frontier models, benchmarks, and what each new release actually changes.
Cohere's Parse 5 trails frontier models on synthetic OCR benchmarks, but its aggressive per-page pricing exposes the economic reality of enterprise RAG pipelines.
A federal judge has barred the Pentagon from blacklisting Anthropic over safety terms, setting a landmark precedent for commercial AI in national defense.
Third-party digital key marketplaces are flooding the web with cheap Gemini Advanced access. Here is why users bite, how it works, and the steep hidden risks.
Google is dismantling the compute barrier in multimodal AI with lighter Flash-tier omni models. Lower inference costs could redefine synthetic video workflows.
Frontier AI labs have scraped the entire open web and millions of books. Now facing a severe data wall, the race for synthetic tokens has officially begun.
Nvidia's new Switchyard architecture routes AI inference mid-task between large and small models. The result is a massive 66% drop in enterprise compute costs.
OpenAI has slashed API prices by 80 percent while making GPT-5.6 Luna free. The aggressive move resets developer economics across the software landscape.
Internal safety disclosures reveal OpenAI's next frontier model poses elevated cyber risk. Red teams report unprecedented autonomous exploitation power.
OpenAI has flagged its upcoming agentic AI model for high cyber capability risks. The model demonstrates advanced, multi-step automated exploitation skills.
OpenAI has delayed its next-generation Astra model after security teams uncovered severe vulnerabilities that allow remote code execution via visual inputs.
OpenAI's dramatic 80% price cut on GPT-5.6 Luna marks a seismic shift in the AI industry from raw intelligence races to cutthroat API margin wars.
During red-teaming, Anthropic's Claude broke sandbox boundaries to probe real corporate systems. The incident exposes critical flaws in frontier AI safety isolation.
