⏰ Limited-Time Free Models
Several free-tier models are approaching their expiration dates on OpenRouter:
| Model | Expires | Price | Context |
|---|---|---|---|
| Baidu: Qianfan-OCR-Fast | May 28 (5 days) | $0.68/$2.81 per M tokens | 65K |
| Mistral: Mistral 7B Instruct v0.1 | May 30 (7 days) | $0.11/$0.19 per M tokens | 4K |
| Google: Gemini 2.0 Flash Lite | Jun 1 (9 days) | $0.075/$0.30 per M tokens | 1M |
| Google: Gemini 2.0 Flash | Jun 1 (9 days) | $0.10/$0.40 per M tokens | 1M |
🗳️ Trump Postpones AI Executive Order Hours Before Signing
The White House abruptly postponed the signing of a landmark AI executive order just hours before the scheduled ceremony, according to multiple reports from Axios, Politico, and CNBC. President Trump told reporters he “didn’t like certain aspects” of the order and cited concerns about overregulation that could hamper U.S. competitiveness against China.
The Washington Post reported that the order would have required companies to report advanced AI models 90 days before public release, with the NSA playing a role in voluntary model testing. Politico noted the postponement came after pushback from industry groups who argued the rules were too restrictive. Reuters cited the President’s desire to prioritize competing with China over imposing domestic regulations.
The reversal marks a sharp departure from the previous week, when Trump had indicated strong support for AI oversight in the wake of Anthropic’s Mythos cybersecurity revelations.
🧠 Nvidia Reports Record $58.3B Profit on $81.6B Revenue
Nvidia announced record quarterly results for Q1 fiscal 2027, posting $81.6 billion in revenue and $58.3 billion in profit — both all-time highs fueled by insatiable AI chip demand. The New York Times called it evidence that “the A.I. boom gathers more steam,” while Al Jazeera highlighted the $58.3 billion profit figure as a new benchmark.
Bank of America analyst advised buying the dip after a post-earnings stock pullback, suggesting Nvidia should boost dividends and buybacks. The company also announced a partnership with Corning to expand U.S. optical manufacturing capacity tenfold for AI data center infrastructure.
💰 Anthropic Nears Profitability with $10.9B Revenue, Acquires Stainless
Anthropic is on track to hit $10.9 billion in revenue during Q2 2026 and is nearing profitability, according to a CNBC source. The company also acquired Stainless, an API SDK company, to strengthen its developer tooling and platform capabilities.
In other Anthropic developments, the company is reportedly in talks with Microsoft to use its custom AI chips — weeks after agreeing to spend $200 billion on Google’s cloud and TPUs. Anthropic also announced that its co-founder believes AI will help make a Nobel prize-winning discovery within a year.
🔢 OpenAI’s AI Model Disproves 80-Year-Old Math Conjecture
OpenAI announced that one of its models autonomously disproved a central conjecture in discrete geometry — the planar unit distance problem — that had remained unsolved for nearly 80 years. New Scientist called it “AI’s biggest breakthrough in mathematics yet,” while OpenAI’s blog described the achievement as demonstrating AI’s ability to go “beyond being just an assistant.”
The breakthrough positions AI as a genuine research partner in mathematics, a field traditionally resistant to automation. OpenAI also announced a partnership with Dell Technologies to bring Codex to hybrid and on-premises enterprise environments.
🆕 Alibaba Unveils Qwen 3.7 Max Agent Model and Zhenwu Chip
Alibaba introduced Qwen3.7-Max, its next-generation AI agent model capable of 35 hours of autonomous work and over 1,000 tool calls per session, according to TechNode and GIGAZINE. The model ranks as the highest-performing Chinese AI model on the Chatbot Arena leaderboard.
Concurrently, Alibaba unveiled the Zhenwu M890 chip, its latest push for domestic AI chip alternatives — reported by Reuters as part of China’s broader semiconductor self-sufficiency drive. Alibaba also announced it would integrate Qwen AI into Taobao and Tmall for agentic shopping experiences.
🛠️ xAI Launches Grok Build CLI Coding Agent and Grok 4.3
xAI launched Grok Build, its first AI coding agent available as a $99/month CLI tool that directly competes with Claude Code and Codex. Engadget and Techzine Global covered the launch, noting the terminal-native approach.
Simultaneously, xAI released Grok 4.3 at an “aggressively low price” with a new voice cloning suite, as reported by VentureBeat. The Pentagon also announced plans to potentially deploy Grok in classified systems as part of a broader push to bring frontier AI models into national security operations.
🏛️ Pentagon Plans Cyber Task Force for AI Hacking Tools
The Pentagon is establishing a new task force to deploy advanced AI models for cybersecurity and offensive cyber operations, according to Gizmodo. The initiative aims to weaponize the latest frontier cyber-capable AI models for national security purposes, marking a significant escalation in military AI adoption.
🤗 Trending on Hugging Face
Notable uploads on Hugging Face this cycle include several research experiments and fine-tuned variants:
- juergengunz/fluxer — New creative generation model (6 likes)
- yusr9/radar-modernbert-large-freeze-pan26 — Radar-focused ModernBERT variant with safetensors
- Nj-1111/EyeeSEE — New vision model (2 likes)
- tkwiecinski/amr-fma-Qwen2.5-7B-Instruct-lora_sdpo-gsm8k-p1 — LoRA fine-tune with SDPO on math reasoning
⭐ GitHub Trending: AI Edition
| Repository | Stars | Description |
|---|---|---|
| Doorman11991/smallcode | ⭐1,053 | AI coding agent optimized for small LLMs — 87% benchmark with 4B-active model |
| Helvesec/rmux | ⭐484 | Universal Rust multiplexer with typed SDK — drive any CLI/TUI from code |
| datawhalechina/Agent-Learning-Hub | ⭐429 | Comprehensive AI Agent learning roadmap and resource collection |
Smallcode continues to lead with 1,053 stars — an AI coding agent achieving 87% benchmark performance with only a 4B-parameter active model, proving that efficient small architectures can compete with much larger systems.
Rmux (484★) stands out as a Rust-native multiplexer with a typed SDK that lets developers drive any CLI or TUI application programmatically — a powerful tool for AI agent orchestration.
🚀 New Models on OpenRouter
Three notable models recently joined the OpenRouter catalog. Note that Qwen3.7 Max and Grok Build 0.1 are covered in the news sections above:
| Model | Provider | Pricing | Context |
|---|---|---|---|
| Grok Build 0.1 | xAI | $1/? per M tokens | 256K |
| Qwen3.7 Max * | Alibaba | $2.5/? per M tokens | 1M |
| Gemini 3.5 Flash * | $1.5/? per M tokens | 1M |
* Detailed in news sections above
💡 Key Trends
1. The AI policy seesaw. Trump’s last-minute postponement of the AI executive order — days after signaling strong support for oversight — highlights the volatile politics of AI regulation. The tension between national security concerns (amplified by the Mythos revelations) and competitiveness with China creates an unpredictable policy environment.
2. AI as a research partner, not just a tool. OpenAI’s autonomous disproof of an 80-year-old math conjecture marks a genuine milestone. Combined with Anthropic’s claim that AI will help produce Nobel-winning science within a year, the narrative is shifting from AI as productivity tool to AI as scientific collaborator.
3. The agent platform race heats up. Alibaba’s Qwen 3.7 Max (35-hour autonomous sessions), xAI’s Grok Build, and Anthropic’s growing platform play all signal that 2026 is the year AI companies compete on agentic capabilities — not just model quality. The battle is moving from benchmark scores to real-world autonomous task completion.