@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.
While sandboxes are a commodity, serving millions of them under production-grade traffic requires attention to detail to scale agents for businesses and consumers. Perplexity is open-sourcing SPACE, the sandbox platform that powers all of Perplexity Computer, to support the ecosystem. https://github.com/perplexityai/space
Why it matters: The distinction between commodity sandboxes and operating millions under production traffic gives context for the engineering demands of scaling agents for businesses and consumers.
Clément Delangue reacts with “Meta is back!” and links to https://huggingface.co/meta-models/Muse-Spark-Open-816B-A42B. In the quoted post, @finkd says Muse Spark Open weights are releasing today and that it runs great on new Mac Studios thanks to work with @johnternus and Apple. https://twitter.com/finkd/status/2110778025629347999
Meet Ollama co-founder and CEO @jmorgan, who speaks about open model adoption in the enterprise. Watch the episode 👇 https://twitter.com/ThisWeeknAI/status/2110754933574082643
Meta announced Muse Spark Open, calling it the most capable US open-weight model to date and describing it as built for coding and agentic work. Meta lists 816B total / 42B active parameters, a 1M token context window, and native text, image, video, and audio support. Ollama welcomed the release and said the model will be available on Ollama shortly. https://twitter.com/AIatMeta/status/2110779754274292212
The same story is shown in the picks as 《Meta releases Muse Spark Open, an open-weight multimodal mixture-of-experts model》
We're open sourcing SPACE, the sandbox platform that runs every task in Perplexity Computer. SPACE gives each agent task its own Firecracker microVM and controls credentials and outbound traffic from outside the VM. Available now under Apache 2.0: https://t.co/DzSmq6o1xA
The same story is shown in the picks as 《Perplexity open-sources SPACE, the sandbox platform powering Perplexity Computer》
Today Meta is releasing Muse Spark Open, which it calls the most capable US open-weight model to date. Built for coding and agentic work, Meta says it excels at long-horizon tasks, complex software engineering, and multi-step reasoning. Key model specs include 816B total / 42B active parameters, a 1M token context window, and native multimodality across text, image, video, and audio.
The same story is shown in the picks as 《Meta releases Muse Spark Open, an open-weight multimodal mixture-of-experts model》
Anthropic has given a team from @METR_Evals employee-like access as part of its commitment to having evaluators assess its safety practices. It is in discussions with multiple organizations. Anthropic describes how it works with third-party evaluators here: https://www.anthropic.com/news/an-update-on-independent-evaluators
Meta’s Muse Spark Open is an open-weight mixture-of-experts model with approximately 816 billion parameters, around 42 billion active per token, and a context window of up to one million tokens. Trained for agentic coding, it accepts text, images, audio, and video and generates text. Hugging Face provides inference and deployment recipes, reports its own quantization tests, and demonstrates attention-layer LoRA fine-tuning with TRL while keeping the experts frozen.
Why it matters: The deployment recipes clarify how sparse expert activation reduces per-token computation while retaining substantial memory requirements, helping developers assess hardware needs across checkpoint formats and serving frameworks.
NVIDIA launches RTX Spark on Friday, Oct. 16, with events around the world where attendees can try hands-on demos for creating, developing and gaming. Events are planned in San Francisco, Berlin, Taipei, Beijing, Istanbul and Seoul; entry requirements vary by location. NVIDIA also invites people to share what their first Windows PC sparked for a chance to win an ASUS ProArt P14 RTX Spark laptop signed by Jensen Huang.
Google is positioning Gemini 4 Argon as a single model for robotics simulation, control, and training data evaluation. Google reports that simulation-generated training episodes raised one robot’s success rate from 56 to 75 percent, but Argon ranked third in Artificial Analysis’s real robot control benchmark. The published demo prompts include extensive guidance, and Google does not report individual demo failure rates or the number of attempts behind the 75 percent result.
Black Forest Labs has released FLUX 3 Image Klein, a 5.6-billion-parameter open-weights model for text-to-image generation, image editing, and generation with up to ten reference images. It supports native 4K output, and its INT8 build runs on NVIDIA GPUs with 8 GB of VRAM, generating an image in under 9 seconds on an RTX 4070. The FLUX Kommunity License allows free commercial use for companies making under $5 million annually.
Why it matters: The hardware requirements and commercial license threshold provide concrete context for assessing whether local image generation fits a team's available GPUs and business size.
Nvidia’s Nemotron 3.5 Super VL is now the leading US open weight model, scoring 26 on the Artificial Analysis Intelligence Index. Artificial Analysis reports twice the intelligence of its predecessor, Nemotron 3 Super (13), using ~60% fewer output tokens per task. https://t.co/VMIiXLcLDj
It's fashion week! https://t.co/o2kkLJ8AAc
We now have state-of-the-art pre-trained base models in the 3T-parameter weight class, built entirely in-house at MAI. Post-training results coming soon—stay tuned! Lots of details in our technical blog post. https://microsoft.ai/news/frontier-pre-training-at-microsoft-ai/
Why it matters: The announcement places MAI’s reported progress at the pre-training stage, with post-training results still forthcoming, providing context for interpreting its claim of state-of-the-art base models.
🎉 Day-0 support for NVIDIA Nemotron 3.5 Super VL in vLLM! 120B total parameters, 12B active per token, and a 1M-token context window. Text, images, documents and video in one open multimodal model, with reasoning and tool use for agent workflows. Here is what is supported and how to get started. 🧵 1/4
vLLM highlights @RedHat_AI’s FP8 and DSpark checkpoints for NVIDIA Nemotron 3.5 Super VL, a 120B multimodal MoE, using LLM Compressor and Speculators. RedHat_AI reports close to 2x performance on Hopper with accuracy intact. Its 4× H100 benchmarks show FP8 improving throughput under load and DSpark boosting per-user streaming speed, with both together outperforming either alone across HumanEval, math reasoning, and RAG workloads. Both checkpoints are open: https://huggingface.co/RedHatAI/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-FP8 https://huggingface.co/RedHatAI/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B.dspark
Plus up your furniture with 3D-printed accessories using Gemini 4 Argon. Upload a pic with rough measurements, and use Canvas in Gemini to build an interactive 3D CAD studio where you can explore styles, tweak live sliders, and export printable STLs to upgrade your space.
The FTC sent civil investigative demands to Anthropic and OpenAI on Wednesday over potential consumer harms from AI safety incidents, an insider told Axios. The requests reportedly exceeded 50 pages and sought incident records, internal security measures, and potentially misleading marketing materials. The source said targeted third parties would also receive subpoenas, and evidence could support litigation if the FTC decides to sue.
Google DeepMind describes how biologists used Google Antigravity to find a novel antimicrobial peptide candidate in *Streptococcus pseudopneumoniae*. ProtNLM helps scan millions of previously uncharacterized UniProt proteins, while the agent coordinates database searches, structural analysis, and literature synthesis. The candidate has predicted activity against pathogens involved in respiratory tract infections and is undergoing wet-lab validation.
Janus Rose examines how Meta’s Muse can collect sensitive information about users and their contacts, raising consent concerns for people who never installed the assistant. Privacy experts argue that permission interfaces obscure the scope of data sharing, while Meta says access is opt-in. The article also explains how uploading messages after decryption can undermine the privacy protections encrypted communications are intended to provide.
Microsoft Research presents Magnetar, a zero-knowledge virtual machine that proves RISC-V program execution using security based only on standard hash functions. It supports programs written in Rust, Go, or C without requiring developers to rewrite them as cryptographic circuits. Across 15 benchmark programs, Microsoft reports average speeds 10 to 39 times faster than the fastest existing comparable zkVMs on the same machine, with proofs verified in about 12 milliseconds.
Why it matters: The benchmark comparisons connect proving speed with general-purpose execution and standard hash functions, clarifying the tradeoffs that have limited broader adoption of verifiable computing.
Nvidia released Nemotron 3.5 Super VL, an open multimodal model with 120 billion parameters and 12 billion active per token, for continuously running enterprise agents. Nvidia says it approaches Nemotron 3 Ultra’s accuracy on specified agent tasks with about one-fifth the parameters, and is releasing weights, datasets, and training recipes. The accompanying Nemotron 3.5 Bolt 74B-A9B 24GB targets local agents, while Bolt E4B 4GB is planned for later this month.
Why it matters: The deployment details clarify how quantization, caching, and hardware requirements shape the practical tradeoffs of running continuous enterprise agents on company infrastructure.
The sudden collapse of the value of CoT (even summarized CoT) in the agentic era changes working with AI. Effectively steering AIs and getting any sort of observability now requires you to ask for interim outputs & tests, and to use AI help to troubleshoot problems you find.
I truly believe conversational AI will drive the biggest educational leap of our generation. 1-on-1 tutoring is the gold standard for education, but it doesn't scale. AI can change that with truly active conversational learning. Annie and team set out to solve this with
you see, meta superintelligence labs is in South Bay so if someone you don’t know is trying to talk to you, you KNOW something is up https://t.co/QQH7Rm78SC
1/ 🚀 New on the vLLM blog: Finish RL Training Faster — Bitwise-Consistent, High-Performance, Agent-Driven Same weights don't guarantee the same policy: inference optimizations can change the logprobs that training recomputes. We make rollout and training bitwise identical https://t.co/S0xkMKfSmc
vLLM-Omni now supports FLUX 3 Image Klein from day 0. The vLLM team worked with Black Forest Labs and NVIDIA to bring the new image model to its serving stack. The integration includes FLUX 3 Image Klein image generation and BF16, FP8 and NVFP4 checkpoints. https://t.co/hsjdzMrbkt
See a fully-explorable 3D world build itself in minutes ⛩️ Watch @GoogleDeepMind's @GeokenAI explain how Gemini 4 Argon acts as a virtual builder in @antigravity to autonomously create a stunning, multi-tiered pagoda sanctuary.
Meet NVIDIA Nemotron 3.5 Super VL. NVIDIA’s new open model combines reasoning and tool use with text, image, document, and video understanding for long-running AI agents. 120B total parameters with 12B active.
The same story is shown in the picks as 《NVIDIA Expands Nemotron Family With Super VL and Bolt Open Models》
Google is rolling out AI labels in the YouTube Music app and website for songs created entirely or significantly altered with AI. Labels appear on the player page and open more information when tapped. Google says it will use partners and its own detection tools to identify AI music, while publishers can optionally disclose whether tracks use fully, partially, or no generative AI.
Black Forest Labs’ FLUX.3 Image Klein launches with day-one Diffusers support for text-to-image generation, multi-reference image-to-image generation, and bounding-box layout control. Open checkpoints include a distilled model using 16 steps without CFG and a base model using 50 steps with CFG; the distilled model generated a 1024 × 1024 image in 1.93 s on an H100 in the reported benchmark. Hugging Face provides quantized loading examples, LoRA training scripts, and workflows for customizing its Modular Diffusers pipeline.
Why it matters: The runtime measurements and quantization guidance help developers assess hardware requirements, including how reference images increase generation time and when lower precision reduces memory without improving speed.
Nilay Patel interviews Benn Jordan about his investigations into Flock Safety and the bipartisan backlash against license plate surveillance. Jordan argues that the central problem extends beyond cameras to private networks that let police search people’s movements, and describes security vulnerabilities he uncovered. He advocates public, immutable audit logs, stronger oversight, and community input while explaining why he now reports vulnerabilities directly to Flock and engages with its leadership.
The Verge reports that AI-generated celebrity images and videos circulate widely on Instagram, sometimes misleading fans, gossip outlets, and Google’s AI overviews. A fake image of Mia Goth drew nearly 500,000 likes before Instagram removed the account for impersonation. Meta exempts still images from mandatory AI labeling, while YouTube offers likeness detection and says it removed over 95 percent of videos reported for takedown through that program in the first half of 2026.
Microsoft has optimized Windows for Nvidia’s Arm-based RTX Spark laptops through easier AI agent installation, tuned Prism emulation, and unified memory improvements. Tom Warren found Hermes setup mostly took two clicks, while OpenClaw required reinstallation and large local models responded more slowly than cloud models. Copilot Plus PC features, including Recall, are unavailable at launch, with initial Windows Insider availability expected later this year.
Mistral AI released Timestral in four sizes, from 5M to 209M parameters, for probabilistic, multivariate forecasting with past-only and known-future covariates. Mistral reports that its 209M model ranks first among pretrained, zero-shot models across five metrics on GIFT-Eval, TIME and fev-bench, using a single checkpoint without ensembling. All sizes support CPU inference, and Mistral published a GitHub repository for running Timestral on users’ data; domain adaptation and fine-tuning are available through Mistral Forge.
Why it matters: The comparisons across model sizes and CPU latency provide a concrete basis for assessing forecasting accuracy against compute requirements in production deployments.
AWS’s Shawn Bice explains Trust in Depth, a framework for governing agents through four layers: Scope, Know, Bound, and Observe. He describes how AWS tools support policies, agent identities, external enforcement, and tracing, using employee offboarding to illustrate sequence-dependent safeguards. Automated reasoning can prove precisely stated rules within a system’s defined assumptions, while context-dependent expectations require evaluation and human judgment.
Why it matters: The offboarding example offers a transferable method for translating business requirements into enforced prerequisites, showing how action order can matter as much as individual tool permissions.
NVIDIA released Nemotron 3.5 Super VL, an open 120-billion-parameter model for multimodal agents, and Nemotron 3.5 Bolt 74B-A9B 24GB for local agentic workloads; Bolt E4B 4GB is coming Oct. 23. Super VL targets tool use, business workflows and visual document understanding, with open weights, datasets and training recipes. Bolt 74B-A9B uses native 2-bit quantization to fit in 24GB of memory, while Bolt E4B targets physical and conversational agents.
Why it matters: Open weights, datasets and training recipes give organizations a path to adapting agents to their own workflows while meeting local deployment and data residency requirements.
