1. WHITE HOUSE LIFTS BAN ON ANTHROPIC'S FABLE 5 AND MYTHOS 5 MODELS
// no sources
// STORY_02
After a 19-day shutdown ordered by the US Commerce Department over national security concerns, the Trump administration lifted export controls on Anthropic's most powerful AI models, Fable 5 and Mythos 5. Anthropic has agreed to give designated federal agencies early access to frontier models before public release and to share threat intelligence going forward.
2. GLM-5.2 TOPS OPEN-WEIGHT RANKINGS, BEATS CLAUDE CODE IN SECURITY BENCHMARKS
// no sources
// STORY_05
Z.ai's open-weight GLM-5.2 model (MIT license, 1-million-token context) scored 62.1% on SWE-bench Pro, beating GPT-5.5 on two of three coding benchmarks, and outperformed Claude Code in Semgrep's IDOR vulnerability detection benchmark. Independent analyses from Semgrep and Graphistry found GLM-5.2 matches or exceeds Claude Opus 4.8 on cybersecurity tasks, with pricing at $1.40 input / $4.40 output per million tokens - roughly one-sixth the cost of GPT-5.5.
3. TOGETHER AI RAISES $800M AT $8.3B VALUATION TO SCALE OPEN-SOURCE INFERENCE
// no sources
// STORY_08
Together AI closed an $800 million Series C led by Aramco Ventures, with NVIDIA, Vista Equity Partners, and General Catalyst participating, valuing the company at $8.3 billion. The startup claims its infrastructure can cut inference costs by up to 60x compared to closed-model providers by running open-source models at scale.
4. PORTUGAL LAUNCHES FIRST OPEN-SOURCE AI MODEL "AMALIA"
// no sources
// STORY_11
Portugal released Amalia, its first sovereign open-source large language model, with training dataset and source code published under an open-source license. Initial applications include a virtual guide for Portuguese museums. The release is part of Europe's broader push for AI sovereignty, reducing dependence on US and Chinese frontier model providers.
5. ANTHROPIC LAUNCHES CLAUDE SONNET 5 AS CHEAPER AGENTIC DEFAULT
// no sources
// STORY_14
Anthropic shipped Claude Sonnet 5 on June 30 as the new default model for all Free and Pro plans, with near-Opus agentic performance at lower API cost. Introductory pricing is set at $3/$15 per million tokens (input/output). Microsoft also made Sonnet 5 available in Azure Foundry for enterprise coding and agent workloads.
6. META PLANS AI CLOUD INFRASTRUCTURE SERVICE, SHAKING UP CHIP STOCKS
// no sources
// STORY_17
A report that Meta is building a cloud infrastructure business to sell excess AI compute triggered a sharp sell-off across AI chip and infrastructure stocks. Meta shares jumped 9% on the news. The move positions Meta as a potential competitor to CoreWeave and hyperscaler cloud providers, with the company having already purchased millions of AI chips from NVIDIA, AMD, and Google.
7. US COMMERCE DEPARTMENT IN ADVANCED TALKS FOR VOLUNTARY AI MODEL STANDARDS
// no sources
// STORY_20
The US Commerce Department is in advanced negotiations with AI companies to establish voluntary standards for new model releases, addressing national security concerns about misuse by foreign adversaries. An announcement could come as early as next week, according to the Financial Times. The standards aim to create a framework for reviewing frontier model releases before public deployment.