Feed

Latest Intelligence

Every AI development we have covered, newest first. Filter by section to focus on what matters to you.

The DecoderOct 5, 2026Enterprise

Aleph Alpha releases Kolibri, an open-weight model that makes the case for European AI sovereignty

Aleph Alpha has launched Kolibri, an open-weight mixture-of-experts model totaling 78 billion parameters with approximately three billion active parameters per token. The German-English model was trained across Germany and Finland on 768 B200 GPUs, with German text comprising more than 21 percent of its training dataset. The model weights are accessible under an Apache 2.0 open-source license.

43/100

The DecoderOct 4, 2026Business

Google's new Gemini tiers cut free users to its weakest model and lock $5/month subscribers out of Pro

Google will restructure access tiers for its Gemini AI services starting in October 2026. Unpaid users will be restricted to the smallest Flash-Lite model, while Flash and Pro tiers will require paid subscriptions, with $5-per-month subscribers excluded from Pro access.

40/100

The DecoderOct 2, 2026Enterprise

Cloudflare says its new Clef model means humans no longer need to be in the loop for AI agents

Cloudflare has introduced Clef and Clef-flash, open-weight models designed to let AI agents make structured decisions without generating text. Built on Qwen and released under the Apache 2.0 license, Cloudflare claims Clef-flash provides classifications in approximately 39 milliseconds, positioning the release as a direct competitor to TypeSafe AI's Jev decision model.

45/100

OpenAIOct 2, 2026Enterprise

A model guide for the GPT-6 family

OpenAI has published a practical guide for the GPT-6 model family, detailing how startups and developers can select models, tune reasoning effort parameters, refine prompts and skills, coordinate tools, and prepare workflows for production.

46/100

The InformationOct 2, 2026Business

How China’s Token Resellers Create an Anthropic Gray Market

According to reporting by The Information, an active gray market of token resellers in Beijing is providing Chinese customers with unauthorized access to Anthropic's Claude and other U.S. artificial intelligence models. Although Anthropic restricts its services in China citing national security concerns, local vendors are circumventing geographic controls to satisfy substantial domestic demand. Anthropic has previously raised concerns regarding Chinese laboratories extracting capabilities from its models.

53/100

CNBC TechOct 2, 2026Business

Google has been playing catch up with OpenAI and Anthropic. Does its new flagship model really compete at the frontier?

CNBC reports that industry analysts view Google's flagship model, Gemini 4, as returning the company to the frontier AI conversation after trailing competitors OpenAI and Anthropic for much of 2026.

41/100

SiliconANGLEOct 2, 2026Enterprise

Microsoft targets ultra-realistic voice agents with its first streaming transcription model

Microsoft expanded its MAI model lineup with its first streaming transcription model alongside two new text-to-speech models. The releases are designed for developers building real-time voice agents capable of listening and responding instantly.

40/100

The DecoderOct 1, 2026Enterprise

Nearly half of test subjects mistook Tavus' AI video avatar for a real person on a one-minute call

Tavus has introduced Griffin, an AI model designed for real-time video calls that processes facial expressions, tone of voice, and gestures. According to a company-run study, 48 percent of test subjects believed Griffin was a real person during a one-minute call, compared to a maximum of two percent reported for prior systems.

40/100

The VergeOct 1, 2026Business

OpenAI’s new agent is a shot at Meta — but can it compete with free?

At its DevDay conference, OpenAI announced Dots, an AI agent powered by GPT-6 Astra, according to The Verge. The product is positioned to compete against Meta's rival Muse AI agent platform as competition among agent ecosystems intensifies.

59/100

The DecoderOct 1, 2026Enterprise

OpenAI says it stopped a campaign to steal its models' reasoning, but the trick still worked on Azure

OpenAI reported stopping a coordinated campaign involving over 15,000 accounts attempting to extract the hidden reasoning processes of its models, linking part of the effort to individuals connected to Moonshot AI. However, researchers discovered that the extraction technique remained functional on Microsoft Azure for weeks, including against GPT-6 Astra, indicating OpenAI's defensive measures were not mirrored on third-party cloud hosting platforms.

71/100