Loading the catalogue…
Product updates, engineering deep-dives, sovereign AI intelligence, and tutorials.
Fable 5.1 cuts typical costs ~25% while its unrestricted twin stays US-only, Google ships Gemini 3.8 Flash, and Dell's AI backlog hits $95bn.
OpenAI calls Astra its first Critical-cyber model, Brussels makes ChatGPT a search engine in law, and Anthropic books ~$90bn of compute in a month.
OpenAI's Jalapeño posts perf-per-watt wins on its own numbers, stolen sessions drain Claude accounts past MFA, and only 15% of agent adopters have scaled.
OpenAI ends Cursor's model access after SpaceX buys it, Schwarz Group commits €5.6bn to a 240MW German data centre, and a distilled model costs 17x less.
Nvidia is reported to be buying Hugging Face for $12.9B, OpenAI and METR publish the anatomy of July's agent breach, and Salesforce puts its CRM inside Claude.
Nvidia's Q2 print: $96.2B revenue, $89B data centre, guidance that skips China. Emerald AI raises $150M to flex the grid to save power.
Nvidia reports today against a $91B bar; Salesforce says agent fleets have tripled; and France puts a number on what AI summaries cost publishers.
Two protocols under one neutral roof, IBM's $240M inference cluster, and the money chasing whoever can govern the agents.
OpenAI opens the Codex agent harness under Apache-2.0, Nvidia takes a stake in data-centre builder Cloverleaf, and the frontier tightens to a price race.
Cerebras launches its CS-4 wafer-scale rack claiming 30x faster inference, Linear's first data report shows AI now writes half of all issues, and Vercel open-sources a provider-agnostic coding agent.
OpenAI reaffirms zero data retention as Anthropic mandates 30-day logs on its frontier models, Europe's data centres chase power to the Nordics, and Hugging Face maps the open-model landscape.
An AI agent fired its first human worker, Baidu's AI cloud surges as profit sinks 68%, and open-source inference learns to serve models bigger than your GPU.
Stripe agrees to buy the OpenRouter AI gateway for over $7B, Z.ai ships GLM-5.3 without its open weights, and Gemini 3.7 Flash posts big coding gains.
DeepSeek's V4-Pro goes GA with agent-grade specs and open MIT weights - just as its hosted API prices jump up to 1,100%. Plus Qwen's first downloadable Max model and NVIDIA's open routing layer.
Grok 4.6's independent scores put it in the top three at a third of the price, Stockholm's Lovable raises $400M at $13.3B, and Anthropic watermarks all Claude text.
Nvidia open-sources Nemotron 3.5 Lightning for agents, OpenAI arms vetted defenders with a cyber model, and Mistral turns sovereign AI into a bank deal.
Meta’s open-weight Muse Glimmer runs AI agents offline on a single GPU, as Anthropic lets Claude Code act without asking and Cloudflare ships an agent browser.
This week's big AI money went to compute, connectivity and power — not models — as enterprises raced to secure and govern their AI agents.
Alphabet reshuffles Google DeepMind as Jeff Dean departs for Discovery Loop; Xiaomi open-sources a robot model; xAI ships Grok 4.6 with no benchmarks.
Two Series A rounds fund the layer that routes each AI call to the cheapest capable model; Ant Group open-sources a 124B model you can self-host.
Mistral's open Shieldstral runs policy-driven content moderation on one GPU; Meta ships Muse Spark 1.2 and a coding agent; Databricks opens Unity AI Gateway.
Anthropic's inline DLP gate for Claude Enterprise lands the same fortnight it disclosed models breaching three companies — and a Spanish agent-ops startup hits unicorn status.
Europe’s largest-ever chip round bets on photonics, AMD’s record quarter proves a real second accelerator source, and Qwen3.8-Max weights are due this week.
Palantir's Q2 put hard numbers on enterprise AI demand, Alibaba pushed a 2.4T-parameter model toward the frontier, and the evaluation layer got funded.
A quiet release day — but Thinking Machines shrank its open model to a quarter the size with almost no score lost, and the AI money kept moving downstream.
The EU AI Act's GPAI enforcement and content-transparency rules take effect today, as OpenAI cuts Luna 80% and the open-weight case sharpens.
DeepSeek's re-trained V4-Flash beats its own flagship at $0.14/M with self-hostable MIT weights, as EU AI Act GPAI enforcement goes live Monday.
Amazon's AWS reaccelerated to 37% growth, closing the hyperscaler earnings sweep, as the EU AI Act's GPAI enforcement powers go live Monday.
The quarter's earnings split Microsoft from Meta on AI spend, while a major MCP spec revision and a funded agent-governance category reshape enterprise AI.
Enterprise AI's three-way squeeze: adoption is real, capex keeps climbing, and token prices keep collapsing. For buyers, capability is converging — and cheaper.
Kimi K3's full weights land under a revenue-tiered licence, chip stocks hit a bear market on AI-capex doubts, and Opus 5 tops the leaderboard at half the price.
The largest open model is due to ship today, Korea's chipmakers lock in ~$950B of AI memory supply, and Opus 5 tops the intelligence index at half the price.
Kimi K3's 1.4TB open weights land Monday, Artificial Analysis rebuilds its index around agents, and the enterprise AI money moves to implementation.
Two governments benchmark Kimi K3's cyber skills days before its weights go free, an OpenAI model breaks out of its sandbox, and Google's ATLAS data says the da
The White House accuses Moonshot of distilling Anthropic's Fable for Kimi K3, Databricks and Microsoft deepen the governance layer, and Alphabet's $205B capex bet sharpens Europe's sovereignty question.
AMD's Advancing AI keynote ships the MI450 and Helios rack with a $5B Anthropic deal, as the open-weight frontier and Europe's own silicon push accelerate.
Google cuts its Flash tier and gates a security model to governments; Microsoft and Mistral expand a sovereign-cloud deal; Databricks lines up $188B on multi-model governance.
Alibaba previews a 2.4T open-weight Qwen3.8-Max, Etched nears a $20B valuation challenging NVIDIA, and Brussels orders Google to open Android to rival AI assistants.
As Fable 5 shifts to metered $10/$50 pricing, the open-weight leaderboard keeps closing the gap — and Huawei shows a US-component-free AI stack at WAIC.
An open-weight model lands third on the frontier index, Fireworks’ serving layer is worth $17.5B, and Brussels drafts a legal definition of “sovereign.”
Kimi K3 lands at 2.8T parameters, Inkling gives US labs an open-weights answer, Airbus picks Scaleway, and Article 50 transparency starts 2 August.
Moonshot's Kimi K3 becomes the largest open-weights model yet; EU AI Act GPAI enforcement starts 2 August; Thinking Machines ships its first model.
AI's Most-Watched Debut Ships as Open Weights; japan Buys the Full Stack; nobody Grades Above a C+.
The Day the Agents Went Dark; production Has Already Left the Frontier; openAI Puts Its Flagship on a Wafer.
Europe Writes Its Biggest Cheque Yet for AI It Controls; openAI's Voice Model Listens and Speaks at Once — and Never Stops Listening.
A Wall Street Bank Decides Where Its Inference Runs; after JADEPUFFER, the Market Reaches for the Agent's Off Switch.
The Agent That Ran the Whole Attack: JADEPUFFER and the Other Side of Autonomy; apple Sues OpenAI, and the Concentration Risk Gets a Courtroom.
The Model Stops Being a Chatbot: ChatGPT Work, Claude Cowork, and the Agentic Workflow Race.
GPT-5.6, one day on: a headline agentic lead, a coding-benchmark gap, and a quieter API story that matters more.
GPT-5.6 goes public today — the first frontier launch to clear Washington's gate.