Google AI Releases Gemini 3.5 Transcribe: A Speech-to-Text Model Reporting 2.6% Average WER Across 85+ Languages

Google's Project Suncatcher satellite returns first AI output from a data-center-class TPU in orbit: onboard Gemma 4 writes a haiku; twin interconnected satellites next in 2027

Google Cloud unveils Agent Compute Engine, the microVM platform behind Gemini agents: one hardware-isolated computer per agent, paused while it thinks, with no credentials inside; preview in the coming months

Google open-sources TorchTPU, a native PyTorch stack for TPUs co-developed with Meta: Transformers, vLLM and SGLang run on TPUs, and Google reports up to 52% MFU across 8,192 Ironwood chips

At Research@ 2026, Google opens Deep Review to the public: an agentic checker for papers and technical documents, piloted on 16,000+ conference submissions; Earth AI agent API enters early access

Google publishes Gemini 4 Argon Safety and Security Report: three-layer real-time defenses, misalignment monitor, Fairwind expansion; 12/12 cyber sims, 0.7% IPI ASR

Google opens Gemini 4 Argon globally: paid Gemini API, Antigravity for Ultra, Gemini app and Search AI Mode for Pro/Ultra; intro price $2/$10 per million tokens