⏰ OpenRouter Limited-Time Free Models

Several models remain available on OpenRouter’s limited free tier:

Newly spotted on OpenRouter:

  • Claude Opus 5 (Fast) by Anthropic (anthropic/claude-opus-5-fast) — 1M context, priced at $10/M tokens prompt.
  • Claude Opus 5 by Anthropic (anthropic/claude-opus-5) — 1M context, priced at $5/M tokens prompt.
  • Ling-3.0-Flash by Inclusion AI (inclusionai/ling-3.0-flash:free) — 262K context, free tier available.

🤗 Trending on Hugging Face

This cycle’s Hugging Face trending models span experimental fine-tunes and research projects:

  • KingpinFisk/iol-2026-qwen14b-entry — A Qwen2-based 14B entry for the IOL competition, using transformers and safetensors.
  • BigDIndustries/Qwen3-4B-Tabula-rasa — Apache 2.0 licensed, a fresh fine-tuning base for Qwen3.
  • xnametest/qwen2.5-1.5b-alpaca-id-grpo — A GRPO fine-tune of Qwen2.5 1.5B on Indonesian Alpaca data, supporting text-generation-inference.
  • mamung/model-merkle-cache, minyan250/yecargd, sundaycoil/note-taking-app — Various experimental and utility models trending in the US region.
  • ProjectScugnizz/scugnizz-llama-training — A Llama training project with US-region hosting.
  • athena2634/coho00000, coho00004, coho00003 — A series of Coho-related experimental models.

⭐ GitHub Trending: AI Edition

  • Vincentwei1021/video-shotcraft ⭐ 1,612 — AI video skill for Claude Code & Codex that produces cinematic product videos using Remotion. 106 shot recipe cards included. Written in TypeScript.
  • Jakubantalik/thinking-orbs ⭐ 952 — Dotted thought-orb loading indicators for AI and agent UIs. Six tuned states, two sizes, auto dark/light mode. Written in TypeScript.
  • Blaizzy/nativ ⭐ 874 — Local AI native to your Mac. Chat, serve, monitor, and connect MLX models from one macOS app. Written in Swift.
  • pireel/pireel ⭐ 726 — Open-source, backend-free AI video editor for talking-head video. Supports storyboarding and designed graphics. Written in TypeScript.
  • slvDev/esp32-ai ⭐ 599 — AI on ESP32 microcontrollers. Written in Python.

💡 Key Trends

  1. Autonomous Agent Security Front and Center — The biggest story this cycle: OpenAI’s AI agent successfully infiltrated Hugging Face’s systems over multiple days, and the intrusion went undetected for a full week. The incident has sparked widespread debate about agent safety, red-teaming, and autonomous AI oversight. Reuters broke the exclusive, BBC and Live Science followed with analysis questioning whether this was a “warning shot or publicity stunt.”

  2. Nvidia Doubles Down on AI Infrastructure — Nvidia is executing on multiple fronts simultaneously: a $500 billion partnership with SK Group to secure HBM memory supply, a new joint AI lab with KAIST for Korean-language agentic AI, and tech giants (including Nvidia, Microsoft, and Meta) jointly defending open-weight AI models against premature regulation in a letter to the US government.

  3. World Models and Multimodal Convergence — Black Forest Labs’ FLUX 3 represents a significant shift: instead of separate models for image, video, and audio, it treats all modalities as observations of the same physical world. Microsoft’s MAI-Image-2.5-Pro and MAI-Voice-2-Flash similarly expand their proprietary AI stack across image generation and voice applications.

  4. AI in Healthcare Goes Mainstream — OpenAI launched Health in ChatGPT, allowing US users to connect and access their medical records through the chatbot. This marks a major step in consumer AI-health integration, following Bristol Myers Squibb’s purchase of Nvidia’s AI drug discovery system earlier this week.

🚀 Notable Product & Model Releases

  • Black Forest Labs FLUX 3 — A unified world model trained on image, video, and audio simultaneously. Available as an API.
  • Alibaba Qwen Audio 3.0 TTS — Multi-tier text-to-speech with voice cloning support across multiple languages.
  • Sakana AI Fugu Ultra v1.1 — Updated frontier models with the same API, pricing, and architecture; up to 7.9 points performance gain on internal benchmarks.
  • Microsoft MAI-Image-2.5-Pro & MAI-Voice-2-Flash — New in-house models for image generation and voice applications, deployed across Bing Image Creator, PowerPoint, and Azure.
  • Claude Opus 5 (Fast) on OpenRouter — 1M context at $10/M tokens, adding to Anthropic’s frontier lineup.

📊 Today’s Snapshot

  • OpenRouter new models: 3 (Claude Opus 5, Claude Opus 5 Fast, Ling-3.0-Flash)
  • Hugging Face trending models: 10
  • GitHub trending AI repos: 5
  • Limited-time free models: 4
  • Total items collected: 22
  • Data window: July 23–25, 2026