← All updates

Gemini 4 Argon Gets Its Wider Rollout

Plus: Meta Ships First Frontier Model, Anthropic Targets Claude Copycats

Good Morning, Model Builders!

Google finally ships Gemini 4 Argon, and Meta answers with Muse Wonder, its first frontier model.


Gemini 4 Argon Gets Its Wider Rollout

TL;DR: Google is expanding Gemini 4 Argon beyond its September debut, rolling it out to AI Pro and Ultra subscribers in the Gemini app and Search AI Mode. Developers can also now access it through the paid Gemini API, while Antigravity remains Ultra-only. Introductory API pricing starts at $2/$10 per million input/output tokens through November. Read more →


Meta Ships First Frontier Model, Muse Wonder

TL;DR: Meta has released Muse Wonder, its first frontier model. Where Muse Spark targeted low latency, the Wonder models are larger and post higher benchmark scores and better results on complex tasks. It gives Meta a credible entry at the top tier and adds one more frontier option for builders to test. Read more →


Anthropic Targets Claude Copycats

TL;DR: Anthropic says it has cut the extraction of Claude's internal reasoning tokens by 90 to 99 percent. The company ties large-scale distillation attacks to Chinese labs Zhipu and Moonshot. Expect tighter controls on what reasoning output is exposed through the API as labs work to protect their models from copying. Read more →


Prime Intellect Acquires Moondream Vision Models

TL;DR: Prime Intellect has acquired Moondream, bringing its open vision models and inference optimizations into the Prime Intellect stack. The combined platform aims to train and deploy multimodal models across vision, video and voice, giving developers of small, efficient open models a more complete home for training and serving. Read more →


Anthropic Prepares Fresh Revenue Disclosure

TL;DR: Anthropic will share updated financials with investors gathering at its headquarters this week, according to people familiar with the matter. As frontier lab financials become central to the tech trade, the numbers will signal how fast paid model usage is growing and how much room the company has to keep spending on compute. Read more →


Anthropic Warns Small Models Could Self-Replicate

TL;DR: A new Anthropic study warns that small AI models may be only months away from spreading on their own across poorly secured consumer devices such as smart TVs and robot vacuums. The research raises the stakes for how open and small models are released, and for securing the devices that could end up running them. Read more →


DeepSeek Eyes Robotics Startup Investment

TL;DR: DeepSeek reportedly plans to lead a funding round for Chinese robotics AI startup Winverse at a $400 million valuation. The move would mark a push by one of China's leading open-weight model labs into physical AI, extending its reach beyond language models into models that run machines. Read more →


AMD Unveils Helion Racks for Agents

TL;DR: AMD unveiled its Helion racks for AI agent workloads at the OCP summit. Partners will ship CPU-only and mixed CPU and GPU configurations from November. The launch reflects a bet that agentic AI will drive demand for CPU-heavy serving alongside GPUs, widening the hardware choices for teams running agents. Read more →


FBI Arrests OpenAI Employee Over Bugging

TL;DR: The FBI arrested a Saudi national working at OpenAI who allegedly planted a recording device in the company's executive conference room. The incident lands as OpenAI and its rivals face heightened scrutiny over their safety and security practices, and as protecting model secrets becomes a growing priority for frontier labs. Read more →


Top AI models. Lower costs.

Access GPT, Claude, Gemini, and more through one OpenAI-compatible API—and save up to 70% on AI API costs. Create your DIT account, get your API key, and start building.