⏰ Limited-Time Free Models

Several limited-time free model offers are available on OpenRouter, with two expiring very soon:

  • NVIDIA: Nemotron Nano 9B V2 — Expires June 11 (1 day left). 131K context. $0.04/M tokens prompt, $0.16/M tokens completion.
  • DeepSeek: R1 Distill Llama 70B — Expires June 11 (1 day left). 131K context. $0.70/M tokens prompt, $0.80/M tokens completion.
  • Z.ai: GLM 4.5 — Expires June 19 (9 days left). 131K context. $0.60/M tokens prompt, $2.20/M tokens completion.
  • Meta: Llama 3 70B Instruct — Expires June 19 (9 days left). 8K context. $0.51/M tokens prompt, $0.74/M tokens completion.
  • Anthropic: Claude Opus 4.6 (Fast) — Expires June 29 (18 days left). 1M context. $30/M tokens prompt, $150/M tokens completion.

🚀 Anthropic Releases Claude Fable 5 — First Public Mythos-Class Model

Anthropic today released Claude Fable 5, the company’s first Mythos-level model accessible to the general public. Described by Fortune as “a version of Mythos the public can access today,” Fable 5 marks a significant departure from Anthropic’s previous policy of keeping its most advanced models behind enterprise paywalls. The model is available immediately on OpenRouter ($10/M tokens input, $40/M tokens output) and directly from Anthropic.

TechCrunch reports that Fable 5 can “make weirdly fun video games with the click of a button,” demonstrating surprising emergent capabilities in creative coding. The release is notable for its timing — coming just days after Anthropic urged AI labs worldwide to pause development, warning that AI systems may soon begin autonomous self-improvement. In a blog post, Anthropic also announced the expansion of Project Glasswing to 150+ organizations, extending Mythos-class model access to research and nonprofit partners.

The company is also investigating a customer data leak following a Friday Claude outage, as reported by Cybernews — raising questions about operational security during this rapid expansion phase.

📱 OpenAI Plans ChatGPT ‘Superapp’ Overhaul Ahead of IPO

OpenAI has confidentially filed its S-1 with the SEC, setting the stage for what could be a $1 trillion market debut. The Financial Times reports that OpenAI is planning a ChatGPT “superapp” overhaul ahead of its listing, aiming to transform the chatbot into a comprehensive platform integrating browsing, voice, video, productivity tools, and third-party agent capabilities.

Added context: SoftBank CEO Masayoshi Son revealed that OpenAI’s next model is being designed by AI — a striking admission about the growing role of automated systems in frontier model development. Meanwhile, Axios reports that Sam Altman dished on OpenAI’s top token user, and OpenAI announced new “Dreaming” memory features for ChatGPT.

An interesting side note: Sam Altman’s eye-scanning company (World) is conducting layoffs, according to TechCrunch — a contrasting data point as OpenAI itself prepares for its public debut.

🇨🇳 China Prepares $295 Billion Nationwide AI Buildout

China is preparing a massive $295 billion plan to fund a nationwide AI infrastructure buildout, according to a Bloomberg News report. The plan, which would be one of the largest coordinated AI investments in history, focuses on building domestic compute capacity, training data centers, and AI chip manufacturing to reduce reliance on foreign hardware.

This comes as Alibaba’s Qwen team opens its AI platform to third-party agents including KFC, Luckin Coffee, and Mixue — signaling China’s aggressive push to embed AI into everyday commerce. Alibaba’s Qwen tech lead also stepped down after a major AI push, suggesting organizational restructuring as competition intensifies.

ByteDance, meanwhile, has set four AI priorities for 2026, with its Doubao assistant losing millions of users — indicating the competitive pressure in China’s domestic AI market.

🦠 Miasma Worm Hits Microsoft — 73 GitHub Repos Compromised

A major supply chain attack dubbed “Miasma Worm” has compromised 73 Microsoft GitHub repositories, including the critical Azure Functions Action repository. The attack specifically targeted AI coding agents — exploiting automated CI/CD pipelines that agents commonly interact with — raising alarms about the security posture of agent-driven development workflows.

StepSecurity reports that the worm disabled repositories through a sophisticated supply chain attack vector. The incident has implications beyond Microsoft, as similar attack patterns could affect GitHub, OpenAI, and Mistral AI infrastructure — the security community is now racing to develop MCP-aware defensive measures.

This event underscores a growing theme in AI security: the very automation tools that accelerate development also create novel attack surfaces.

🏛️ House Unveils AI Draft to Preempt State Laws

The U.S. House of Representatives has unveiled a draft AI bill that would preempt state-level AI regulations, creating a single federal framework. The move comes as states from Florida to Connecticut to Colorado race to enact their own AI laws — Connecticut Governor Lamont just signed comprehensive AI legislation including youth online safety provisions.

The House draft is positioned as a compromise between industry calls for regulatory certainty and consumer protection advocates’ demands for robust oversight. Trump’s administration simultaneously asked AI firms to voluntarily submit models for cybersecurity testing, indicating a preference for voluntary rather than mandatory frameworks.

AI policy adviser Krishnan also announced departure from the White House, adding to turnover in key AI policy positions.

In Brief

  • SoftBank commits €75 billion to French AI data centers — The largest single AI infrastructure deal in European history, announced at the Choose France summit.
  • Nvidia and SK hynix announce multiyear chip partnership — Advancing memory technology for AI factories, alongside new deals with SK Group and Naver for gigawatt-scale AI infrastructure in Korea.
  • Mistral AI lands €722 million to deepen Europe’s AI infrastructure, partnering with Airbus and TCS for sovereign aerospace and enterprise AI applications.
  • Meta unveils Muse Spark — Its first new AI model since hiring Alexandr Wang, designed for smart glasses and on-device AI. API release delayed after performance concerns.
  • Italy closes DeepSeek probe after the company committed to warning users about AI hallucination risks — a regulatory resolution that could set a precedent for other EU member states.
  • Google cuts AI Plus plan price, doubles storage — Engadget reports the price reduction as competition intensifies in the consumer AI subscription market.
  • Arcee AI built a 400B-parameter open-source LLM from scratch — The tiny startup claims to have surpassed Meta’s Llama on several benchmarks, per TechCrunch.
  • Lovable hits $500M annualized revenue — The AI web app builder now processes 1 million new projects per week.

🤗 Trending on Hugging Face

  • sundaycoil/alert-router — An endpoints-compatible model designed for intelligent alert routing in monitoring pipelines, earning early community traction with 2 likes.
  • chilkersion/KurutAI (3 likes) and chilkersion/adondlin255 (1 like) — Continued experimental uploads from the same author.
  • JensLundsgaard/morfea-2026-06-09 (54 downloads) — A ConvLSTM autoencoder using PyTorch model hub, demonstrating HF’s steady role as a research model distribution channel.
  • mulemp/search_bot — A search-focused utility model with no clear pipeline tag.

HF ecosystem activity is relatively quiet today — most uploads are experimental or low-download personal projects rather than major releases.

🚀 New Models on OpenRouter

  • Anthropic: Claude Fable Latest — The newest Mythos-class model from Anthropic, available immediately. 1M context window. $10/M tokens input, $40/M tokens output.
  • Anthropic: Claude Fable 5 — The stable release of Anthropic’s first public Mythos model. Same pricing and context as Fable Latest.
  • Nex AGI: Nex-N2-Pro (free) — A 262K-context model available on OpenRouter’s free tier, giving developers zero-cost access to a capable agent-oriented model.

⭐ GitHub Trending: AI Edition

  • JimLiu/baoyu-design ⭐628 (+86★) — Run Claude Design locally as an Agent Skill for Cursor, Claude Code, and more to produce polished UI mockups. Rapid growth continues.
  • GordenSun/GordenSuperPPTSkills ⭐574 (+113★) — AI-powered PPT generator creating luxury image-format presentations via GPT, converted to editable PPTX. Surge in popularity.
  • tastyeffectco/sandboxd ⭐535 (+19★) — Self-hosted dev sandboxes with preview URLs. One command, no Kubernetes — perfect for coding agents.
  • amElnagdy/guard-skills ⭐510 (+36★) — Quality gates for coding agents that catch AI-generated failure modes in code, tests, and docs.
  • razr001/align-dev ⭐354 (+9★) — Helps AI-assisted frontend teams generate shared coding standards and SKILL.md files for Claude Code.

💡 Key Trends

Today’s news reveals a deepening paradox of AI safety and acceleration: Anthropic releases its most powerful public model while simultaneously calling for a global pause. The supply chain security theme emerges as a critical new dimension — the Miasma worm demonstrates that agent-driven development creates novel attack surfaces. China’s $295B infrastructure plan underscores the geopolitical compute race, where AI dominance increasingly depends on national-level industrial policy rather than company-level R&D. And the OpenAI “superapp” vision signals the next logical phase of AI monetization — platform consolidation around a single AI operating system for daily life.