News✳
The releases, research and policy decisions shaping AI around the world. Original context, with sources you can follow.

Genie 4 adds sound to interactive AI worlds: access, limits and what to watch
Google DeepMind’s October 27 research preview generates worlds with sound and live scene steering. Broader access is still ahead, and continuity remains a research problem.

GPT-6.1 Terra launches: access, pricing and what to test before automating screen work
OpenAI’s October 26 release targets visual reasoning, computer use and writing. Here are the access details, API requirements and pricing conditions behind the announcement.

Mercury 3 arrives: launch pricing, API access and the limits of a speed headline
Inception’s October 19 release promises faster text generation. Current documentation lists discounted API rates, while the practical question is how much waiting disappears from a complete workflow.

NVIDIA releases Bolt E4B: what its 4GB target means for local AI assistants
The October 23 Nemotron release targets local conversations and tool use. Its memory measurements show why developers should check the workload, checkpoint and full application together.

Kimi K3.1 Pro launches: access, API pricing and when Highspeed is worth testing
Moonshot’s October 23 release brings a faster Highspeed option and a million-token context window. Hosted access is live; full model weights are promised by November 4.

1Password for Agents launches: approved logins, rollout dates and what access still means
The October 22 launch keeps passwords outside an agent’s view. Muse and Grok Bot integrations are rolling out; Claude Duet and ChatGPT Work connections are still listed as coming soon.

Claude Duet arrives: who can use it, usage limits and a useful first test
Anthropic’s October 21 research preview brings background work to Claude. Access starts with US Pro and Max subscribers, and every background check uses the same plan allowance.

Mistral launches Voxtral 3: voice agents, pricing and what to check before taking calls
The October 20 speech-model release connects listening, speaking and business tools. Model cards label it a public preview, and some launch details need checking before a rollout.

MiniMax M3.1 Flash launches: pricing, access and the cost of a finished task
The October 19 release targets coding and AI agents. Here is how to try it, what the API costs and why a cheaper response needs a completion test.

DeepSeek opens its search API: pricing, Flash 1017 and what to test
The October 17 release pairs web search and page retrieval with an updated Flash model. Here is what developers can access, what requests cost and how to assess the results.

Reflection releases Beam: open weights, coding agents and what to test first
Beam became available on October 16. Here is what the release offers developers, how Mirror CLI fits in, and why an open model still needs a careful trial.

Meta Muse Spark Open explained: specs, pricing and what it means for you
Meta’s new open-weight AI offers more control over coding and agent workflows. Here is a practical guide to its capabilities, costs, hardware, license and the tests worth running before you switch.