# News: October 7, 2026

> 9 curated links on AI agents, Python, security, Google Cloud, and engineering, selected by Médéric Hurier.

- Author: [Médéric Hurier (Fmind)](https://www.fmind.dev/)
- Published: 2026-10-07
- URL: https://www.fmind.dev/news/2026-10-07/

- [Mistral Large 4](<https://mistral.ai/news/mistral-large-4>): Mistral AI announces Mistral Large 4, an open-weights natively multimodal mixture-of-experts model with 1.05T total parameters, 49B active parameters, and a 1M token context window in public preview.
- [EmbeddingGemma 2 – An open, lightweight multimodal embedding model](<https://blog.google/innovation-and-ai/technology/developers-tools/embeddinggemma-2>): Google introduces EmbeddingGemma 2, an open lightweight multimodal embedding model designed for privacy-first, on-device retrieval and decision-making.
- [Sharing AI progress in mathematics](<https://openai.com/index/sharing-ai-progress-in-mathematics>): OpenAI shares research preprints highlighting AI-driven discoveries across open problems in mathematics and engages the research community to evaluate automated proofs.
- [Bring multimodal semantic search to the edge with EmbeddingGemma 2](<https://developers.googleblog.com/google-ai-edge-with-embeddinggemma-2>): Google details deploying EmbeddingGemma 2 with MediaPipe and LiteRT to achieve private, ultra-low-latency on-device multimodal semantic search across text, vision, and audio.
- [EmbeddingGemma 2 – The Developer Guide](<https://developers.googleblog.com/embeddinggemma-2-the-developer-guide>): Google publishes a comprehensive guide for building multimodal search and RAG applications with EmbeddingGemma 2 across text, code, images, video, and audio.
- [Wiz AI SAST – AI Code Security With Code-to-Cloud Context](<https://www.wiz.io/blog/introducing-wiz-ai-sast>): Wiz launches Wiz AI SAST in public preview, providing AI code scanning that identifies broken access control and IDOR vulnerabilities while connecting them to cloud risks.
- [Meta and Microsoft scale back internal use of Anthropic's Claude](<https://finance.yahoo.com/technology/ai/articles/meta-microsoft-scale-back-internal-185842649.html>): Meta and Microsoft scale back internal employee use and spend on Anthropic's Claude models, prioritizing in-house tools like GitHub Copilot and MetaCode to control operating costs.
- [Nano Banana on X](<https://x.com/NanoBanana/status/2107521202466046453>): Nano Banana announces Nano Banana 2.1 in the Gemini App and Google AI Studio, delivering upgrades in visual design, mask-based editing, subject consistency, and natural imagery.
- [Mistral is BACK! (Le Chonk)](<https://www.youtube.com/watch?v=Hu1JOK6aXsI>): Matthew Berman breaks down the release of Mistral Large 4, analyzing its 1T-parameter MoE architecture, benchmark performance, and upcoming open-weights timeline.