Last week, several of South Korea's largest banks were hit by a cyberattack. A CrowdStrike report reportedly indicates the entire attack may have been carried out by one person. The attacker reportedly combined the open-source AI penetration tool ARTEX, DeepSeek v4.1-Flash, GLM-5.3, Grok 4.6, and Claude Code.
Preference Model is open-sourcing Karotte, the framework it uses to build reinforcement learning environments that resist reward hacking, including defenses like killing stray processes before grading and rejecting grader-crashing files. The framework has been hardened through more than a million evaluation runs and controlled red-teaming. The company focuses on machine learning engineering tasks for leading labs, and a16z says it is partnering with Preference Model and its founders, Jennifer Zhou and Ning Cao.
Reflection AI and Mistral each unveiled new open-source models this week, aiming to beat other Western open models, though they trail top Chinese and closed systems on prominent benchmarks. Reflection CEO Misha Laskin says the target is regulated industries and governments that cannot or will not use Chinese models. The outcome depends on whether businesses and agencies accept less advanced models for some tasks in exchange for lower cost and more control.
Mistral released Mistral Large 4 "Le Chonk", a 1T-parameter (49B active) multimodal model, with open weights planned in about three weeks. Google rolled out Nano Banana 2.1 across Gemini, AI Studio, and the Gemini API, and released EmbeddingGemma 2, a 740M-parameter open multimodal embedding model under Apache 2.0. OpenAI launched the Decisions API in beta with gpt-6-luna, returning typed answers 10x faster than the Responses API.
OpenAI's Math post is up! https://github.com/openai/math
This demonstration used a JavaScript injection via MathJax rendering, which an agent could trigger from anywhere in its output, including its reasoning. For clarity, we haven't seen agents exploit this in our evaluations, and Inspect was patched within a day of our report.
I have a feeling this is a good open source AI week 🥰 - Monday: @reflection_ai Beam - Tuesday: @MistralAI Mistral Large 4 - Who is next? 👀👀 More is coming this week! 🤩
OpenAI says it will ship one meaningful Codex and Work improvement each day for 28 days starting October 5, or else offer a "reset" without specifying what that reset covers. The company also plans to test visual ads in ChatGPT image generation in the U.S. starting in late October, with ads kept separate from generated images and not affecting answers.
Congrats to the Reflection team on Beam 🎉 Be sure to join us next week at our NYC meet-up with the Reflection team and NVIDIA https://luma.com/TAIxNvidiaxReflection
Reflection is set to release Beam, its first open-weight AI model, aiming to become a Western counterpart to DeepSeek. The source says Beam is trained from scratch for coding, reasoning, and AI agents, with benchmarks placing it alongside the strongest open models and more efficient token economics. Reflection has raised $4.6 billion from investors including Nvidia, Sequoia, and Lightspeed, and the interview covers its monetization plans for open-weight models.
BIG MOMENT FOR AMD🚨THIS WEEK: AMD HAS REACHED ABOVE 90% PARITY ON UPSTREAM vLLM GATING TEST GROUPS! This was after months of hard work from AMD maintainers, specifically Andreas, and also vLLM CI Lead Kevin, along with SemiAnalysis getting vLLM enough CI AMD GPUs after months of upstream maintainer complaints about not having enough AMD GPUs on vLLM and AMD CI fleet-wide stability issues. (1/3)🧵
We're enjoying our Saturday morning ☕, a bit of 📰 and seeing our AMD vLLM commits. Open source FTW.
Keras is moving to a pluggable backend design, with MLX and PaddlePaddle backends upcoming as add-on libraries. The team is reducing the operations needed to ship new backends and streamlining unit testing so a single harness can test all ops, such as casting consistency. KerasHub also gains many new models and is shifting its preprocessing from tf-text to PyGrain.
Goodfire says AI cybersecurity risks are already here and that biosecurity risks are next, as models improve at biology. The post presents this as both an opportunity for science and medicine and a reason for stronger security. It quotes Demis Hassabis announcing SynthID for biology, a watermarking approach for AI-generated proteins, published in Nature with SynthID Bio tools open sourced.
Nice to see @FireworksAI_HQ leading the open model inference economy 🎆
ESM-2 came out in 2022. It's still downloaded hundreds of thousands of times a month. That only works because someone keeps maintaining the software underneath it. @huggingface 🤝 @os4science are teaming up to find those libraries and back the people behind them 🧬 https://os4science.org/news/hugging-face-open-source-for-science-fund/
Lovable's security team found a vulnerability (CVE-2026-102989) in TanStack Start, which allows attackers to run unwanted JavaScript in visitors' browsers via crafted links. Lovable reported it to TanStack and deployed firewall protections for hosted apps while a fix was prepared, and affected projects will be automatically updated on their next change or via the Security page. Lovable says it found no evidence of exploitation in reviewed logs, and apps hosted elsewhere must apply the upstream update themselves.
We're co-hosting an open source party with @HuggingFace! 🤗🦥 Come join us, we'll be handing out @UnslothAI merch, showcasing our upcoming features and more! It'll be a super fun night with lots of demos, DJs, food and more. Join: http://luma.com/OpenTogether Use code: NOSLOTHSHERE to get in.
🤯🤯🤯🤯 10:46 sign in with ChatGPT, which lets people sign into your app and use the tokens they are paying for already.
Our Secure Intelligence Institute works with researchers at Stanford, CMU, Duke, Columbia, Ohio State, and UVA. We’re also working with NVIDIA and the Open Secure AI Alliance, and sharing tools and findings so others can strengthen their own systems. https://blogs.nvidia.com/blog/open-secure-ai-alliance/?ncid=so-twit-957725
Anthropic reports that Zhipu AI's GLM-5.3 can autonomously build end-to-end cyber exploits and is released without meaningful safeguards against misuse. In its simulated tests, attackers bypassed the model's safeguards 64% to 100% of the time using simple techniques, while the same attacks failed against safeguarded Claude models. Anthropic also cites an NIST CAISI assessment calling GLM-5.3 the most cyber-capable open-weight model released to date.
AIWhy it matters: The report shows how open-weight safeguards fail under simple bypasses, offering concrete test figures for judging misuse risk in released models.
数字生命卡兹克 announced that AIHOT, an AI hotspot news site with about one million monthly active users, is now open source on GitHub. The release includes the collection pipeline, curation scoring, clustering mechanism, and the production prompts, aiming to let others build vertical versions for industries such as gaming, law, HR, and finance.
Today Microsoft announced Autopilot, an always on agent built on OpenClaw The best part of this collaboration is how much @OmarShahine and others at Microsoft have contributed BACK to OpenClaw Read all about the contributions Omar and his team made here: https://openclaw.ai/blog/microsoft-autopilot-openclaw
In 7 months, the skills.sh registry reached 1 million agent skills and nearly 280 million installs. See what people are teaching agents, and where skills are going ↓ https://vercel.com/blog/state-of-agent-skills
With @GoogleDeepMind, @emblebi, and research partners, we’re making AI-predicted protein complex structures for 2,800+ viruses openly available. This gives scientists a head start in preparing for potential outbreaks.
inclusionAI has published public training-content summaries on Hugging Face for its Ling and Ring model versions, including Ling-2.0, Ling-2.5, Ling-2.6-1T, Ling-3.0, Ring-2.0, Ring-2.5-1T, and Ring-2.6-1T. The documents, organized under the template associated with Article 53(1)(d) of Regulation (EU) 2024/1689, contain documentation only, not model weights or training datasets. Each summary covers only the model versions it names.
Mike Knoop says the top ARC Prize 2026 ARC-AGI-2 score of 83.06% by Tufa Labs is only 2% short of the 85% grand prize threshold. The challenge runs under strict Kaggle compute limits with no internet access, and the winning solution is set to be open sourced. The image shows the leaderboard with RabbitHole at 76.94%, nvbanana at 74.17%, Yi-Chia Chen at 55.14%, and Kha Vo at 37.50%.
Thanks @ValsAI for the high-caliber eval! The term "flash" is a bit "misleading" now. With 124B total size and 5.1B activation, Ling-3.0-flash-fin is a "flash lite" with high intelligence density. Enjoy the free API while it last. We also have fp4 quant to be used on local AI 😛
Jev pod tomorrow subscribe on apple /youtube @latentspacepod thanks to @allenpark and Ke for making this happen!
Can't wait to see some of you in Berlin!
Exciting results, @LiamFedus! Congrats to the whole team at Periodic Labs!
Google says it is focusing AI efforts on health, disaster and weather resilience, learning, and economic opportunity. Recent examples include AlphaGenome Atlas, which maps all 9B possible single-letter genetic changes across the human genome and is openly available to researchers, and WeatherNext 3, described as its most accurate and capable global weather AI model to date. The post also cites AI & Economy ATLAS, an open-access look at global AI usage, and says its translation services now cover nearly 300 languages spoken by 7B people.
Good timing by DeepSeek to release just after we removed support for seq2seq models in TRL 😅
🌐 Supporting open source. Expanding deployment options. We’ll work closely with the open-source community on V4.1-Flash inference support and explore more deployment options. Planning a large-scale deployment with 2,000 GPUs + a storage cluster? Let’s talk. 🔹 Model: https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash 🔹 Paper: https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash/blob/main/DeepSeek_V41_Tech_Report.pdf 6/6
Qwen3.8-27B Unsloth GGUF is now the #1 most-liked GGUF of all time! The model hit 10M downloads and 3.7K likes in just 24 days on Hugging Face - all thanks to you. 🤗🦥 GGUF: https://huggingface.co/unsloth/Qwen3.8-27B-GGUF Guide: https://unsloth.ai/docs/models/qwen3.8
Ali Ghodsi congratulated Clement Delangue and Jensen Huang and said the deal will be good for the industry and open source. The author's own post is short, and the quoted NVIDIA message says NVIDIA is acquiring Hugging Face and will be its new home.
Today, we are launching Tinker grants of up to $50,000 in credits for safety research on open-weight models. We share some project ideas that excite us below; if you’re working on a safety project that could be accelerated by additional Tinker credits, we want to hear from you!
Z.ai says GLM-5.3 is its most capable model for cybersecurity tasks, with CyberGym at 84.5% versus 77.2% for GLM-5.2 and ExploitBench at 54.4% versus 24.4%. Access will begin with selected security partners in controlled settings, followed by broader access and API availability, with full open weights to be published after safety evaluations are complete. The company also launched the OpenVuln initiative to help open-source maintainers audit projects and coordinate disclosure.
Meta announced it is opening the weights for Muse Glimmer, a 30B parameter dense model that can run locally. Muse Spark 1.2, described as its latest foundation model, will have its weights released soon. The author's interview with Mark Zuckerberg quotes him saying Llama 4 fell short of the trajectory he wanted and that the lab was rebuilt.
Thank you Mark, Alex and the whole Meta team for your contributions to open weight AI.