New models and versions, open weights, and changes in capability and pricing, with their evaluations.
Aravind Srinivas announced two open-sourced Decision Models: pplx-decider-flash-v1.2: 4B params, 1 cent per million tokens, and the top 4B model on the HF Decision Index. pplx-decider-v1.2: 27B params, 2 cents per million tokens, and #2 overall on the HF Decision Index.
Jeff Dean writes, “Have a glimpse at this! Nice work, @AndrewDai and Elorian folks!” https://twitter.com/AndrewDai/status/2111142161584628060 In the quoted announcement, Andrew Dai introduces Glimpse as their first visual thinking model, claiming competitive frontier performance compared with Astra on key public visual reasoning benchmarks.
Google is rolling out Gemini 4 Argon globally to developers, enterprises, and paid subscribers, emphasizing coding, professional knowledge work, multimodal understanding, and sustained task execution. API pricing is $2 per million input tokens and $10 per million output tokens through November 30, rising to $4 and $20 on December 1. Google reports that Argon worked unattended for 11 hours on ADK optimizations, with engineers merging two changes; Ethan Mollick described improvements but noted remaining weaknesses in complex multi-agent tasks.
UBTECH introduces Thinker-WM, a multimodal world model that jointly models RGB video, depth, and robot actions through a shared Unified Dynamics Transformer. In material supplied by UBTECH and republished by QbitAI, the company reports leading results on WorldOlympiad, EWMBench, and WorldArena1.0 Track-1, including scores of 0.740 on WorldOlympiad and 4.66 on EWMBench. The model supports six capability modes through shared weights, and adding depth improved RoboTwin single-task success by 15 percentage points over a baseline without depth.
Elorian released Glimpse on Friday as a research preview with API access and a waitlist-only chat interface. Its visual scratch space lets it crop image regions, retrieve video frames, and run code to measure or compare visual content during reasoning. Pricing is $0.60 per million input tokens and $3 per million output tokens; Elorian’s evaluations report performance above GPT-6 Astra on scientific chart reasoning and short-form video captioning.
NVIDIA Nemotron 3.5 Super VL is now available on Ollama. The model combines reasoning and visual understanding across a range of agent workflows. Ollama says it delivers up to 5x higher throughput on a single GPU for agentic tasks compared with other open models of its size.
Meta has launched Muse Spark Open, a natively multimodal open-weight model with 816 billion parameters and a 1 million token context window, built for complex coding tasks and agentic workflows. It is designed to run locally on Macs, though Engadget reports it requires at least 1 terabyte of RAM. Meta claims superior performance over other popular open-weight models in its evaluations and is making the model available now through multiple API platforms.
The same story is shown in the picks as 《Meta releases Muse Spark Open, an open-weight multimodal mixture-of-experts model》
Baoyu reports that Meta released Muse Spark Open on October 15, an open-weight model with 816B total parameters, 42B active parameters, and a 1M token context window. It accepts text, images, video, and audio and outputs text; the Muse Model License permits free commercial use subject to attribution, usage restrictions, and a separate agreement for qualifying API businesses. The model card lists minimum memory requirements of 700GB for NVFP4 and 2TB for BF16, while Meta describes optimization for two Mac Studio machines with 512GB unified memory each.
The same story is shown in the picks as 《Meta releases Muse Spark Open, an open-weight multimodal mixture-of-experts model》
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
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》
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’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
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.
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》
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.
Project Apertus released Apertus 1.5 in 8B and 70B sizes, adding image and audio understanding, optional thinking, a 262,144-token context window, and improved instruction-following and tool use. The weights were released on Hugging Face on July 24, 2026 under Apache 2.0; the models generate only text, and spoken-language processing remains experimental. The team also publishes training data, recipes, intermediate checkpoints, and reproducible evaluation records.
Why it matters: Published intermediate checkpoints, training mixtures, and evaluation settings provide a concrete basis for investigating how individual training stages change model behavior and for reproducing reported results.
Skywork AI released Matrix-Game 4.0, an open-source game world model that reportedly runs at 480p and 60 FPS on one RTX 5090, with input response under 0.15 seconds and peak VRAM usage of approximately 18.5GiB. A game engine maintains state and executes rules, while a video model generates visuals from structural frames and dynamic captions. Model weights and code are available through https://huggingface.co/Skywork/Matrix-Game-4.0 and https://github.com/SkyworkAI/Matrix-Game.
The same story is shown in the picks as 《Kunlun Tech releases and open-sources Matrix-Game 4.0 for single-GPU real-time gameplay》
Huawei says its mass-produced Kirin 9050 Pro uses logic folding to shorten signal paths through vertically connected circuit layers, supporting performance and efficiency gains in the Mate 90 Pro Max. The article explains how dense hybrid bonding enables this layout and how memory coordination allows Kirin 9050 chips to run a 30B MoE model with less than 2.5GB of additional memory usage. Huawei also says Kirin 2027 chips are in testing, with results meeting expectations so far.
@itsnoahd says Muse Wonder (not full) comes in 4th on Noahbench, just below Opus 5.5, and calls its price-to-performance-to-speed ratio “perfect.” They say each task takes 10 minutes. Alexandr Wang calls it “a chic entry into noahbench.” https://t.co/VLAlRoBsDN
Gemini 4 Argon taught a robot to play cornhole just by watching a video. It mapped the physics, coded its own simulation, and taught @Apptronik's Apollo humanoid to sink 7 throws in a row.
Why it matters: The demonstration connects video observation, physics modeling, and a model-written simulation in a concrete example of teaching a humanoid robot a physical task.
Google DeepMind has launched Gemini 4 Argon globally through its API and selected Google products, with access depending on subscription or service tier. Introductory pricing is $2 per million input tokens and $10 per million output tokens, with cached input tokens at 95% off and a 1M unified input-output token limit through the Gemini API. Google reports improvements across coding, professional workflows, scientific research, and robotics, alongside stronger agent safeguards and prompt injection resistance.
Why it matters: Google’s reported gains in code submissions and reductions in rollbacks connect Argon’s engineering capabilities to delivery speed and code stability in internal workflows.
Gemini 4 Argon is rolling out today to Google Ultra and Pro subscribers in @GeminiApp and Search, and to Ultra subscribers in @antigravity. Demis Hassabis praises its knowledge work and coding capabilities, sharing a wizard chess game they asked it to create as an example of its physics understanding and 3D modeling capabilities that reminds him of Battlechess on the Amiga. More information and full availability: https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/
The same story is shown in the picks as 《Gemini 4 Argon is now available worldwide as Google's most capable frontier model》
Meta released Muse Wonder and Muse Wonder Full on October 12, according to Baoyu’s account of the announcement. Developers can access Muse Wonder through Meta Model API or Muse Code, while Full is available to Muse Code subscribers; Meta’s blog and Zuckerberg’s X post differ on consumer availability. The models support a 1 million Token context window and multimodal input with text output, while Meta’s reported benchmarks show strengths in sustained agentic and coding tasks alongside weaker results elsewhere.
The same story is shown in the picks as 《Meta releases frontier models Muse Wonder and Muse Wonder Full》
Le Chonk cooked on visual grounding. Pinpoints exactly where things are in an image. 42% on Dense 200, ahead of even frontier closed models..
vLLM congratulated the vllm-sr team on releasing Decision 3.0, described as a state-of-the-art multimodal Jev model whose weights are now open-sourced on Hugging Face. The quoted announcement from Xunzhuo Liu says the open model family spans 0.6B to 27B and ranks #1 on both Jev Decision Index boards in its evaluation, on a new Pareto frontier.
Odyssey has launched a free public research preview of Odyssey-3, a world model that generates interactive environments from text prompts and runs in real time on an autoregressive diffusion transformer. The base model has 14 billion parameters and outputs 832 × 480 video, while Odyssey-3 Pro supports 1280 × 720. Its reported 66.1-point Physics-IQ Verified score came from a single run using a selection method, below the benchmark's four-run requirement.
Liquid AI released two open-weight models in its d1 decision-model family: d1-3B and the experimental d1-omni-600M.
Why it matters: Liquid AI open-sourced two decision models and reported on-device latency and results across seven benchmarks, providing evidence on the use of small models for structured decisions.
Perplexity is releasing pplx-embed-v2-late, described as two late-interaction embedding models that retrieve text, images, and pages through a shared embedding space for cross-modal querying. The company says both models reach frontier performance and are publicly available on Hugging Face, with a blog post at perplexity.ai/hub/blog/multimodal-embeddings-beyond-a-single-vector.
Google DeepMind released EmbeddingGemma 2, a 740M-parameter open model based on the Gemma 4 architecture and licensed under Apache 2.0. It maps text, images, audio, and video into a shared embedding space.
Why it matters: EmbeddingGemma 2 maps text, images, audio, and video into one embedding space and reports concrete on-device memory and vector truncation figures, helping assess local multimodal retrieval.
Google released EmbeddingGemma 2, an open multimodal embedding model that converts text, images, video, audio, and code into numerical vectors. It has 740 million parameters. Google says it outperforms competitors with more than twice as many parameters on multimodal embedding benchmarks and scores 78.68 on the Massive Text Embedding Benchmark (Code), nearly 10 points above the previous model's 68.76.
Simon Willison used the same prompt to ask Mistral Large 4, claude-opus-5.5, gpt-6.1-sol, and gemini-3.8-flash each to generate an SVG of an armadillo jaywalking on Mars in fishnet stockings, comparing frontier models after benchmark saturation. He ran the test with the LLM command-line tool, using each model's default reasoning level.
Google DeepMind released EmbeddingGemma 2, which it calls the first natively multimodal open embedding model for on-device use. It combines code, images, audio, and video with text in a shared space.
Sundar Pichai announced EmbeddingGemma 2, described as Google's first open, natively multimodal embedding model. It handles text, code, image, video, and audio tasks in a modular 740M parameter form factor, is positioned for offline, privacy-first RAG when paired with Gemma 4, and is said to outperform some specialist models more than twice its size. Weights are available now on Hugging Face.
French AI lab Mistral AI released Mistral Large 4 (ML4) on Tuesday, a 1-trillion-parameter multimodal model nicknamed Le Chonk. It is currently accessible only through a guarded public endpoint, with open weights planned within three weeks after safety testing is complete.
Mistral released its largest model to date, Mistral Large 4, with 1 trillion parameters and 49 billion active parameters. An API preview is available in Mistral Studio; weights are expected at the end of October.
Mistral released a public preview of Mistral Large 4, a natively multimodal model with 1 trillion parameters and 49 billion active parameters. Its weights will be opened at the end of the month.
Why it matters: Mistral outlined the open-weight release schedule for its 1-trillion-parameter multimodal model and its investment in European compute, allowing comparison with mainstream open models.
Mistral AI released Mistral Large 4, codenamed Le Chonk, with 1T total parameters, native multimodality, and 49B active parameters. The company says it is the best open-weight model on aggregated benchmarks in Europe and the U.S. It reaches state of the art on workloads in cyber defense, manufacturing, and finance, and exceeds closed frontier models in visual grounding. Built end-to-end in Europe, it can be deployed from Europe through Mistral Cloud. The API is available today, and open weights are due at the end of October.
Why it matters: Mistral specified a 1T-parameter, 49B-active open-weight model and a local European deployment path, offering a view of the open-weight field's scale and availability.
Reka AI released a research preview of Rho-1, a 19B-parameter general-purpose model that can process and generate text, images, video, and robot-control actions in a single neural network. Unlike most systems that route tasks to specialized models, Rho-1 places all modalities as tokens in one shared context window without calling tools or external models. It can generate continuous video in real time, respond instantly to new instructions without restarting, and use the same weights that predict camera images to drive robot actions.
Tavus launched Griffin, which the company calls the first Human Interaction Model (HIM). In live video calls, it can receive and generate speech, facial expressions, tone of voice, gestures, and pauses at the same time.
Google DeepMind released frontier model Gemini 4 Argon, initially making it available through the Fairwind Program to trusted cybersecurity defenders before gradually expanding access. Pricing starts at $2 per million input tokens and $10 per million output tokens.
Why it matters: Gemini 4 Argon raises the maximum output to 1M tokens and reports multiple benchmark results and internal deployment data, offering a view of its limits on long-running workflows.
