⏰ Limited-Time Free Models
Several free-tier models are expiring soon on OpenRouter. Xiaomi’s MiMo-V2-Omni and MiMo-V2-Pro are the most urgent — expiring May 31 (just 2 days away). The Gemini 2.0 Flash line follows on June 1.
| Model | Expires | Pricing (Prompt / Completion) |
|---|---|---|
| Xiaomi MiMo-V2-Omni (262K ctx) | May 31 | $0.40 / $2.00 per M tokens |
| Xiaomi MiMo-V2-Pro (1M ctx) | May 31 | $1.00 / $3.00 per M tokens |
| Google Gemini 2.0 Flash Lite (1M ctx) | Jun 1 | $0.075 / $0.30 per M tokens |
| Google Gemini 2.0 Flash (1M ctx) | Jun 1 | $0.10 / $0.40 per M tokens |
| Qwen3 30B A3B (131K ctx) | Jun 5 | $0.09 / $0.45 per M tokens |
| Llama 3 Euryale 70B v2.1 (8K ctx) | Jun 5 | $1.48 / $1.48 per M tokens |
| Hermes 2 Pro Llama-3 8B (8K ctx) | Jun 5 | $0.14 / $0.14 per M tokens |
| Claude Opus 4.6 Fast (1M ctx) | Jun 29 | $30 / $150 per M tokens |
💸 DeepSeek Makes V4-Pro 75% Price Cut Permanent
DeepSeek has permanently reduced the price of its flagship V4-Pro model by 75%, escalating the AI pricing war to a new level (Engadget, InfoWorld, VentureBeat). The permanent cut follows a temporary promotional period and solidifies DeepSeek’s strategy of aggressive pricing to capture enterprise market share. The move puts pressure on Western labs to respond with competitive pricing of their own, particularly in cost-sensitive enterprise deployments where total inference cost per token is becoming the deciding factor.
🇨🇳 Moonshot AI Raises $2B at $20B+ Valuation, Debuts Kimi K2.6
Chinese AI startup Moonshot AI has closed a $2 billion funding round led by Meituan, valuing the Kimi chatbot maker at over $20 billion (Bloomberg, SCMP, SiliconANGLE). The company simultaneously released Kimi K2.6, a 1-trillion-parameter open-source model featuring:
- Agent swarm scaling — orchestrates up to 300 sub-agents across 4,000 coordinated steps
- Long-horizon coding — capable of running code-generation tasks for days without human intervention
- Attention optimizations that significantly reduce inference cost
The release positions Moonshot as a serious competitor to both OpenAI and Anthropic in the agent-capable model space. Cloudflare has also begun running Kimi K2.5 on its Workers AI platform, signaling growing adoption.
🏭 Infrastructure Boom: ByteDance $70B, Dell +32%, Goldman $800B
The AI infrastructure spending race continues to accelerate:
- ByteDance is considering a staggering $70 billion in capital expenditure for 2026, according to Bloomberg and The Information — a figure that would rival the entire AI capex budget of major US hyperscalers. The spending would go toward AI chips, data centers, and compute infrastructure.
- Dell Technologies stock surged 32% in a single day — its best performance ever — after reporting AI server revenue that dramatically exceeded analyst expectations (CNBC, Reuters). Hewlett Packard Enterprise also rallied on the sector-wide tailwind.
- Goldman Sachs raised its 2026 AI spending estimate to $800 billion, up from earlier projections, citing sustained hyperscaler capex and enterprise adoption (Yahoo Finance, Benzinga).
🔬 Anthropic: Mythos to Finance Watchdog, Samsung/SK Hynix Invest
Anthropic has agreed to share its findings on the Mythos AI model’s cybersecurity vulnerabilities with the Financial Stability Board (FSB) — the global finance watchdog — marking a rare instance of an AI company proactively disclosing cyber threats to financial regulators (The Guardian, NYT). The move comes amid ongoing investigations into unauthorized access to Mythos.
In parallel, Samsung and SK Hynix have joined Anthropic’s $65 billion Series H round, the strategic investment reflecting the growing interdependence between AI model developers and memory/semiconductor manufacturers (KED Global, theinvestor.co.kr). Anthropic’s valuation now stands at $965 billion, surpassing OpenAI to become the world’s most valuable AI company.
🏛️ Colorado Signs First AI Chatbot Regulation for Minors
Colorado Governor Jared Polis has signed new legislation regulating how AI chatbots interact with minors — the first state-level law of its kind in the US (Colorado Politics, The Denver Post). The law requires:
- Age verification mechanisms for chatbot platforms
- Disclosure that users are interacting with AI, not humans
- Limitations on data collection from minors
- Opt-out options for parents
The bill signals a growing wave of state-level AI regulation as federal efforts remain stalled.
🚀 New Models on OpenRouter
Three new models appeared on OpenRouter today:
- StepFun: Step 3.7 Flash — $0.20/M tokens prompt, 256K context. StepFun’s latest efficient flash model.
- Anthropic: Claude Opus 4.8 (Fast) — $10/M tokens prompt, 1M context. Fast variant of the latest Opus iteration.
- Anthropic: Claude Opus 4.8 — $5/M tokens prompt, 1M context. Standard variant with the same massive 1M context window.
⭐ GitHub Trending: AI Edition
| Repository | Stars | Language | Description |
|---|---|---|---|
| study8677/awesome-architecture | 808 ★ | Vue | 21 architecture maps covering AI gateways, RAG, and design patterns |
| UditAkhourii/adhd | 514 ★ | TypeScript | Tree-of-thought coding agent skill with pruning, built on Claude Agent SDK |
| withkynam/vibecode-pro-max-kit | 496 ★ | JavaScript | Spec-driven coding harness for AI-assisted development |
| FlashML-org/flashlib | 390 ★ | Python | Fast and memory-efficient classical ML operators |
| 2aronS/Duel-Agents | 357 ★ | TypeScript | CLI, SDK, and IDE plugins for multi-agent systems |
🤗 Trending on Hugging Face
Hugging Face activity was relatively quiet today. Among the new uploads, a multilingual model checkpoint — cs-552-2026-mnlplus/multilingual_model (183 downloads, Qwen3-based conversational model) — stands out as the only upload with meaningful download activity. Most other new models show zero or near-zero downloads, reflecting a slower day for the open-weight ecosystem.