Cloud and AI news, Nov 12: Google Gemini 4 Hydrogen

Transcript

Gemini 4 Hydrogen is now generally available with a unified limit of 1,048,576 tokens. Plus Batch TPU support, Inkling 2 Small, MiniMax image model and four more, in Thursday's cloud updates.

It adds four thinking levels, implicit caching and Office inputs, plus an xhigh level, at $1.50 per million input tokens and $7.50 per million output.

Google Cloud Batch jobs can now run on TPU7x, v6e and v5p TPU VMs for machine learning and AI workloads.

Over at Cloudflare, Workers AI now offers Inkling 2 Small, a multimodal model with 594 billion parameters and a 1,048,576 token context.

From MiniMax, there is a new open-weight image model, released by the company.

Over at AWS, FSx for Lustre Intelligent-Tiering now costs less to start, with minimum throughput falling from 4,000 to 1,000 MBps.

From Azure, Front Door Premium now supports client certificate authentication with four validation modes and root-CA-only trust configuration.

Over at AWS, Glue can now upgrade Apache Spark jobs to the latest runtime automatically when the version is set to Auto.

From Anthropic, you can now run the full Managed Agents service on your own infrastructure, storing agents and sessions there.

So, that's Thursday. A busy stretch for AI platforms, from new models and TPU support to running agents on your own infrastructure.

Follow for tomorrow's updates. The full list is at cloud changelog dot com.