Viewpoints, commentary, analysis, interviews, and discussion of trends.
@kirillk_web3 says Meta’s Llama 4 disappoints everyone, Meta goes closed with Muse Spark, Zuck writes a $100-200 billion check, Muse Wonder ships on Monday, and 3 days later Meta drops you with open weights: 816 billion parameters, only 42 billion active at a time. https://t.co/031gtR0nHm https://t.co/nEN1JMzvBa Alexandr Wang: “lmao muse spark open balling out” https://t.co/AcUT9o8NEQ
Guys this is like my superbowl I waited so long for US open weight models to be relevant again. Before today the US models were still behind Chinese models from as early as Feb. 2026. Still more work to do!
Guys this is like my superbowl I waited so long for US open weight models to be relevant again. Before today the US models were still behind Chinese models from as early as Feb. 2026. Still more work to do! This cut the gap by at least half, in time (say 3 months)
@AIatMeta says it is releasing Muse Spark Open, which it calls “the most capable US open-weight model to date.” The quoted post says the model is built for coding and agentic work and lists 816B total and 42B active parameters, plus a 1M token [context window](https://t.co/yAtoxNeYJm). Nathan Lambert reacts: “Okay now I really gotta blog 🎉” https://t.co/uKl5X18Xpw
Nathan Lambert predicts adoption of muse spark open will surge if its performance roughly matches its reported scores, calling it the first US open-weight model with near-frontier agentic capabilities. He compares the shift with GLM 5.2 and Kimi K3 in China. In the quoted announcement, @alexandr_wang describes muse spark open as having 816B total params, 42B active, 1M-token context, and native multimodality, built for coding and agentic work. https://t.co/9elY7xigU5
The Meta Muse Open License is trash. Thankfully no one really cares about licenses, expect the inference providers who will happily comply.
Simon Willison highlights indie game developer Popcar’s account of a recent wave of AI-powered game decompilation, enabling unusual game mashups and preservation projects that bring old games to new hardware, often in browsers using WebAssembly. Popcar describes risks including competitors copying software and paid apps being open-sourced against their creators’ wishes, alongside benefits such as easier ports, modifications, and continued access when game servers shut down.
I joined the a16z podcast with @lishali88 to explore how open models fit into RSI dynamics, who captures their value, how to wrangle agent swarms, and why I’m building Trillium Labs. https://t.co/psWUUGCfrS
Very cool to see Embedding Gemma 2 by @Google, number on trending on HF! The media and silicon valley bubble only talks about reasoning LLMs when there is a much bigger variety of AI models that are super impactful and useful like embeddings, decision models, audio, video, time-series,... (over 3M public models on HF!). Let's put them back in the spotlight!
I can't wait to do much much much more in France with @huggingface @nvidia in the coming years. France has the talent & the energy to become the third country in the world for datacenters and AI in general, especially thanks to their strong open-source AI ecosystem. Let's do it 🇫🇷🇫🇷🇫🇷!
Open AI models offer research access, customization, and local deployment, but downloadable weights also make misuse harder to monitor and access impossible to revoke. Grace Huckins examines these trade-offs through expert interviews, including arguments about concentrated power, cybersecurity, and hardware access. The explanation also covers staged releases and tamper-resistant training, noting that current tamper-resistance approaches do not withstand all attacks.
If you’re a researcher, please consider sharing your RL environments openly! Something you built for one experiment could help the entire field move forward. I know we’re in dark days for sharing from frontier labs in the US, but I believe this can have a huge impact on the field and the world!
I've been having a lot of fun trying a wide range of LLM post-training experiments using KerasHub + JAX + TPUs. Might open-source some of that
By end of 2027 I'll be the only open models person Nvidia hasn't acquired
We've been monitoring a decline in HuggingFace downloads across the top open LLMs we track for interconnects. It looks like there was a slight change in counting of downloads, as we saw a ~30% download rate drop for many top models, but this is continuing over time. What else do people think it may be?
MIT researcher Alex Zhang discussed his Recursive Language Models (RLM) on the Latent Space podcast. An RLM-based harness is the first system to come close to solving ARC-AGI-3, ahead of OpenAI's Astra.
OpenAI officially open-sourced Codex Harness on August 19, releasing the Agent Loop behind Codex App, CLI, and IDE Extension. Developers can use the SDK and App Server for task management, tool calls, event callbacks, and human approval; the models and cloud services remain provided by OpenAI.
Of the 45 layers in GLM-5.3-Flash, 34 use Kimi's KDA linear-attention approach and 11 use DeepSeek's KPool-DSA sparse attention. Linear and Sparse attention appear together in one model for the first time. The combination reflects a shift away from finding a single best mechanism toward assigning different roles: Linear maintains long histories at low cost, while Sparse retrieves specific information as needed.
Radical Numerics co-founder and CEO Eric Nguyen argues that biosecurity is an AI arms race: models that increase biological capabilities can also help defenders keep pace with the offensive capabilities of frontier models.
NVIDIA argues that AI safety is an engineering problem requiring enforceable controls, clear ownership, and verification evidence, and says safety boundaries must operate independently of agent reasoning. Its open-source NVIDIA OpenShell safety runtime enforces policies outside an agent's reach and provides sandboxed execution. Cisco DefenseClaw and JFrog have built governance and skill-scanning capabilities on top of it.
Hugging Face has released a report on the open-model ecosystem from January through August 2026, drawing six findings from Hub data. Public model repositories grew from 2.43 million to 2.96 million and datasets from 711,000 to one million, but 85.6% of models have fewer than 200 lifetime downloads, while 1.5% of repositories account for 99.2% of downloads.
Why it matters: Hugging Face analyzes its Hub data on the open-model ecosystem; findings on downloads versus likes and the scale of Qwen derivatives can inform model selection.
We share our approach to open-weights releases, how we assessed Inkling, why safety depends on both the model and the ecosystem it enters, and how testing, staged access, and stronger defenses can create a path toward greater openness.
Releasing weights indiscriminately isn't safe. Neither is keeping capable models inside a few labs. We think there's a path between them. We haven't mapped all of it. Our new post covers the part we can see: how we assessed Inkling, and why access should widen in stages. https://thinkingmachines.ai/blog/a-safe-path-to-open-weights
The knowledge that makes AI useful is diffused. It lives with scientists, engineers, clinicians, firms. For AI to benefit from distributed knowledge, it must itself be distributed. Agree with Jensen that this is a future worth building.
An analysis by the UK's AI Security Institute finds that the gap in cybersecurity capabilities between open-weight models and closed frontier models is narrowing. GLM-5.2 comes closest to Claude Opus 4.6, released 4.3 months earlier; DeepSeek-V4-Pro falls between Claude Opus 4.5 and GPT-5. For much of 2025, the measured gap was six to ten months.
