The latest AI news in plain words: new models, AI tools, research, cloud and policy updates, plus picks from people learning AI. Updated every day, every story linked to its source.
Models ·
Grok 4.7 improves on Grok 4.6 for long-running coding and knowledge tasks, with better self-verification and context management, while keeping pricing at $2 input and $6 output…
SpaceXAI
Research ·
AlphaGenome Atlas provides predicted molecular effects for every possible single-letter change in the human genome, about 9 billion variants, in a dataset of roughly one…
Google DeepMind
Models ·
OpenAI began rolling out GPT-6 Astra, which it calls its most capable and most aligned model, through ChatGPT, the API, Azure and AWS Bedrock, priced at $10 input and $50 output…
OpenAI
Business ·
NVIDIA announced a deal to acquire Hugging Face for $12.93 billion, and pledged that the hub will stay open, multi-cloud and usable without NVIDIA hardware. Hugging Face hosts…
NVIDIA
Models ·
Gemini 3.7 Flash is Google's new workhorse model for coding and agents, with reported gains over 3.6 Flash on software engineering benchmarks such as DeepSWE (65.3% vs 49.0%)…
Models ·
DeepSeek made V4-Pro generally available in its app and API, adding agent upgrades, low/high/max reasoning-effort settings, and native support for the OpenAI Responses API tuned…
DeepSeek
Alibaba unveiled Qwen3.8-Max, a sparse mixture-of-experts model with 2.4 trillion total and about 95B active parameters aimed at coding and 'cowork' agent tasks, and committed to…
Alibaba Qwen
Policy ·
From 2 August 2026, providers and deployers covered by the AI Act must tell people when they are interacting with an AI system and label certain AI-generated content, including…
European Commission
Tools ·
The Model Context Protocol's 2026-07-28 specification removes the initialize handshake and session IDs so each request stands alone, and adds HTTP headers for method and tool…
Model Context Protocol
Moonshot AI published weights for Kimi K3, a 2.8-trillion-parameter mixture-of-experts model with about 104B active parameters, a 1M-token context window and native text, image…
Moonshot AI
Models ·
Claude Opus 5 launched at $5/$25 per million tokens, with Anthropic saying it comes close to Fable 5 on coding benchmarks such as CursorBench at about half the cost per task. The…
Anthropic
Research ·
Hugging Face reported that an autonomous agent system broke into its production infrastructure through its dataset-processing pipeline, and said it found no tampering with public…
Hugging Face
Models ·
OpenAI released GPT-5.6 in three tiers: Sol as the flagship, Terra as a mid-tier model, and Luna as the low-cost option, priced from $5/$30 down to $1/$6 per million tokens at…
OpenAI
Cloud ·
AWS added organizational knowledge integration, web search and a managed knowledge base for retrieval to Amazon Bedrock AgentCore, and launched background agents in Amazon Quick…
AWS
Z.ai released GLM-5.2, a roughly 753B-parameter successor to GLM-5.1 with a 1M-token context window and multiple thinking-effort levels, under an MIT license without regional…
Z.ai
Business ·
SpaceX said it would exercise its option to buy Anysphere, the company behind the Cursor AI code editor, in an all-stock deal valued at $60 billion; the acquisition closed in…
CNBC
Policy ·
The Commerce Department's Bureau of Industry and Security required an export license for any foreign person, including Anthropic's own non-US staff, to access Fable 5 and Mythos…
CSIS
Models ·
Anthropic released two versions of its Mythos-class model: Fable 5 for general use with safeguards, and Mythos 5 with some safeguards lifted for vetted cybersecurity and…
Anthropic
Cloud ·
At Build, Microsoft introduced in-house MAI models including MAI-Code-1-Flash for GitHub Copilot, a GitHub Copilot app in technical preview, and general availability of the…
Microsoft
Models ·
Claude Opus 4.8 launched at the same $5/$25 per million token price as Opus 4.7, with Anthropic reporting it is about four times less likely to miss code flaws. Alongside it…
Anthropic
Business ·
Anthropic closed a $65 billion Series H at a $965 billion post-money valuation, led by Altimeter, Dragoneer, Greenoaks and Sequoia, including $15 billion from hyperscalers. The…
Anthropic
Models ·
At I/O, Google released Gemini 3.5 Flash across the Gemini app, Search AI Mode, Antigravity and the Gemini API, saying it beats Gemini 3.1 Pro on coding and agentic benchmarks…
Policy ·
The Council and European Parliament provisionally agreed on an AI 'omnibus' package that pushes high-risk obligations for stand-alone systems to 2 December 2027 and for AI…
Council of the EU
DeepSeek published open weights for DeepSeek-V4-Pro (1.6T total, 49B active parameters) and V4-Flash (284B total, 13B active), both with 1M-token context and thinking and…
DeepSeek
