Frontier AI
API usage tiers are now Build, Launch and Grow, and organizations move up automatically as total credit purchases reach each tier minimum.
Why it matters. Rate limits and monthly usage limits follow your tier, so check which tier your organization landed in before planning capacity.
Source: OpenAI API changelog
Administrators can turn on both Claude models for Antigravity 2.0, Antigravity CLI and Antigravity for IDEs in the global, us and eu locations.
Why it matters. Third-party models are off by default and bill as consumption with overages enabled, so decide your opt-in policy before teams ask.
Source: Gemini Enterprise release notes
In cloud environments with limited networking, allowed_hosts now also restricts web_search and web_fetch, and mismatched allowed_domains entries fail session creation with a 400 error.
Why it matters. Agents that fetch pages outside allowed_hosts now get url_not_allowed errors, so add those hosts or remove the domains.
Source: Claude Platform release notes
The 753B mixture-of-experts model has a 1-million-token context window, up to 128K output tokens and explicit prompt caching.
Why it matters. Eligible enterprise customers can run an open weight coding model through US and Global cross-Region inference without hosting it.
Source: AWS What's New
The speech-to-speech model improves reasoning, instruction following and tool calling, with a 256K context window and seven languages, at Nova 2 Sonic pricing.
Why it matters. Voice agent teams can upgrade at the same price in four Regions, including Stockholm and Tokyo.
Source: AWS What's New
The 740 million parameter open model embeds text, images, video, audio and code, with an 8K token context and 768 dimensions.
Why it matters. Teams can download it from Hugging Face or Kaggle and run multimodal search on their own hardware with commercial use allowed.
Source: Google blog
The in.moonshotai.kimi-k3 inference profile serves Kimi K3 from the Mumbai and Hyderabad Regions.
Why it matters. Teams that must keep inference inside India can now use Kimi K3 through Bedrock.
Source: Amazon Bedrock document history
Pegasus 1.5 turns video into structured, time-coded metadata, reading picture, sound and on-screen text in one pass without pre-indexing.
Why it matters. Media and compliance teams can search video libraries by moment through US and Global cross-Region inference.
Source: AWS What's New