OpenAI cuts its top model prices by up to 80%, DeepMind ships whole-body robotics, and Anthropic discloses its own AI breached three companies — on a day when Big Tech confirmed $1.1 trillion in AI infrastructure spend since 2023.
OpenAI has cut prices on its GPT-5.6 Luna and Terra models — Luna down 80%, Terra down 20% — continuing the industry-wide compression of frontier model costs.
DeepMind released Gemini Robotics 2 and its Embodied Reasoner 2, claiming the models enable generalist robots to handle complex, unscripted physical tasks without per-task fine-tuning.
Anthropic disclosed that Claude Opus 4.7, Mythos 5, and an unnamed model each successfully breached a real company's defenses during sanctioned red-team security evaluations — details withheld at the companies' request.
The GCC (GNU Compiler Collection) steering committee has released an AI governance policy covering acceptable use of AI-generated contributions in the project.
Across Amazon, Alphabet, Microsoft, and Meta, combined capital expenditure since 2023 has reached $1.1 trillion, with $745 billion more committed for the remainder of 2026 — almost entirely AI infrastructure.
DeepSeek pushed a V4-Flash update to its API, improving latency and throughput on its open-weight frontier model.