Google's Gemma 4 12B lands free and open, Qwen3.7 Plus hits OpenRouter
Google shipped Gemma 4 12B, a multimodal open model you can run today on Hugging Face. Qwen3.7 Plus also appeared on OpenRouter, confirmed across three sources.
Google shipped Gemma 4 12B, a multimodal open model you can run today on Hugging Face. Qwen3.7 Plus also appeared on OpenRouter, confirmed across three sources.
Google just released Gemma 4 12B — a multimodal, encoder-free open model that scored 780 points on Hacker News and landed simultaneously on the Google blog and Hugging Face. If you've been waiting for a capable open model you can actually run locally or fine-tune without a research license, this is the one to look at this week.
Google's Gemma 4 12B is described as a 'unified, encoder-free multimodal model' — meaning it handles text and images in a single architecture without a separate vision encoder bolted on. The instruction-tuned version (gemma-4-12B-it) is already on Hugging Face at rank #5 in trending models. It's cross-confirmed by Hacker News, Hugging Face, and Reddit r/LocalLLaMA. At 12 billion parameters it sits in a sweet spot: powerful enough to be useful, small enough to run on a single consumer GPU with quantization.
Alibaba's Qwen3.7 Plus appeared on OpenRouter today under the handle qwen/qwen3.7-plus. It's cross-confirmed by Hugging Face model listings and Reddit r/LocalLLaMA. OpenRouter gives you API access immediately, so you don't need to self-host. No benchmark numbers were published alongside the listing, so treat capability claims from community posts with caution until independent evals appear.
ComfyUI shipped v0.24.0 on GitHub. If you run image generation workflows locally, this is a routine but real update worth pulling. Cross-confirmed by Hugging Face model activity. No changelog summary was available in the signal data — check the GitHub releases page directly.
llama.cpp tagged build b9499. Given that Gemma 4 12B just dropped, expect a GGUF conversion and llama.cpp support to follow within days if not hours — the community moves fast on new Google releases. Cross-confirmed by Reddit r/LocalLLaMA.
OpenAI published a 'blueprint for democratic governance of frontier AI' and a separate public policy agenda document. Both are position papers, not product launches. The governance blueprint was picked up by Hacker News. Worth reading if you care about the regulatory environment your tools will operate in — but nothing changes for builders today.
OpenAI's blog published a case study on how Wasmer used Codex to build a Node.js runtime for edge computing. Cross-confirmed by Reddit r/ChatGPT. It's a real example of AI-assisted systems programming — useful reading if you're evaluating whether Codex-class tools can handle low-level infrastructure work.
Anthropic published research mapping AI-enabled cyber threats against the MITRE ATT&CK framework, covering a full year of data. If you're building anything that handles sensitive user data, this is worth a skim — it names real attack patterns, not hypothetical ones.
1. Pull gemma-4-12B-it from Hugging Face and run a multimodal test: feed it a product photo and ask it to write a listing description. See if 12B is enough for your use case before paying for API calls.
2. Try Qwen3.7 Plus on OpenRouter's playground today — no self-hosting needed. Compare its output on a coding task against whatever model you're currently using. OpenRouter's free tier lets you test without a credit card.
3. If you run a ComfyUI image workflow, update to v0.24.0 and check whether any of your existing node graphs break — catching that now is faster than debugging it mid-project.
Confirmed: Gemma 4 12B is live on Hugging Face (google/gemma-4-12B-it) and announced on the Google blog. Cross-confirmed by Hacker News and Reddit r/LocalLLaMA.
Confirmed: Qwen3.7 Plus is available on OpenRouter at qwen/qwen3.7-plus. Cross-confirmed by Hugging Face and Reddit r/LocalLLaMA.
Not independently verified by CBW: We have not run benchmarks on Gemma 4 12B or Qwen3.7 Plus. Community benchmark posts should be treated as preliminary until third-party evals appear.
Worth noting: The Qwen3.7 Plus OpenRouter listing included no published benchmark numbers in the signal data. Capability claims circulating on Reddit are not independently verified.
Worth noting: OpenAI's governance blueprint and public policy agenda are position documents. They signal regulatory intent but do not change any product or API today.
Source: Google blog — Gemma 4 12B launch — https://blog.google/innovation-and-ai/technology/developers-tools/introducing-gemma-4-12b/
Source: Hugging Face — google/gemma-4-12B-it — https://huggingface.co/google/gemma-4-12B-it
Source: OpenRouter — qwen/qwen3.7-plus — https://openrouter.ai/qwen/qwen3.7-plus
Source: OpenAI blog — Wasmer and Codex case study — https://openai.com/index/wasmer
Source: Anthropic — AI-enabled cyber threats and MITRE ATT&CK — https://www.anthropic.com/news/AI-enabled-cyber-threats-mitre-attack
Source: GitHub — ComfyUI v0.24.0 — https://github.com/Comfy-Org/ComfyUI
Get tomorrow's best AI project in your email. With a guide that works. Free. No spam.