AI news for : 7 updates. Every card links to the maker's own page.
Clairic is a preview extraction model on Workers AI that returns typed values from a state and schema instead of generating text.
Why it matters
Teams with structured-output tasks can extract invoice totals, delivery dates, shipment delays, or repair costs.
The Cloudflare API MCP server can now request approval for an additional scope when an agent needs it during a task.
Why it matters
Users of compatible MCP clients can choose Minimal at connection time and grant more access only when a tool call requires it.
Vercel AI Gateway now offers Inception's Mercury 3, a diffusion model that generates text through parallel refinement.
Why it matters
Builders of browser agents, real-time search, interactive education, and customer support can use its improved reasoning.
The deprecated models are Gemini 3.7 Flash for Gemini 3.8 Flash, GPT-5.5 and GPT-5.4 for GPT-6.1 Sol, GPT-5.4 mini and GPT-5 mini for GPT-6 Luna, and Grok 4.5 for Grok 4.7.
Why it matters
Copilot Enterprise administrators may need to enable alternative models through model policies in Copilot settings.
The LiteRT Tensor API lets developers build a computation graph operation by operation in C++ and choose how its operations connect.
Why it matters
Developers who need low-level graph control can keep model inference and processing on an accelerator without rewriting the graph for each target.
Amazon Web Services
Assistant changesAmazon Bedrock now supports SpaceXAI Grok 4.7 for coding, agentic tasks, and knowledge work in AWS GovCloud (US).
Why it matters
AWS GovCloud users can start using Grok 4.7 in the Amazon Bedrock console or through supported Amazon Bedrock APIs.
MiniMax
Assistant changesMiniMax M3.1 Flash is now available through the MiniMax API, M Plan, and MiniMax Code.
Why it matters
Teams can apply M3.1 Flash to agentic tasks, coding, computer use, and cybersecurity.
Find dated assistant changes, one table per assistant, in the tracker.