Auto Added by WPeMatico
Reka has released a research preview of Rho-1, a 19B omni-reasoning model trained from scratch. A single neural network understands and generates text, images and video, reasons over them, and outputs robot actions. Reka frames it as a direct replacement for agentic pipelines that pass work between modality-specific models. What Rho-1 Changes Most multimodal systems
Researchers from PhAI Labs, CUHK, Fudan, Stanford, Oxford and Princeton have released JEPA-Anything, a domain-agnostic framework for building world models. Instead of designing a new predictive model for each field, it applies one shared learning recipe to very different systems. It extends joint-embedding predictive architectures (JEPAs) with a method called Orthogonal Predictive Factorization (OPF). The
Beyond Domain-Specific World Models: JEPA-Anything Uses 1 Recipe for 7 Fields Read More »
Together AI has released Together Link, a free, MIT-licensed CLI now in beta. It connects the coding agents developers already use to open models hosted on Together AI. Supported tools include Claude Code, Claude Desktop, Codex, ChatGPT Desktop, OpenCode, and Pi. The idea is simple: keep the harness, swap the model, and shrink the bill.
In this tutorial, we design an end-to-end streaming robotics learning pipeline around the NVIDIA Cosmos3-DROID dataset without downloading its 707 GB repository locally. We first introspect the LeRobotDataset v3.0 structure and construct a metadata graph from info.json, task metadata, episode tables, and dataset statistics, then use HTTP byte-range access with PyArrow to selectively read Parquet
Building a Streaming Robotics Learning Pipeline Using NVIDIA Cosmos3-DROID Read More »
Reflection AI has introduced Beam, its first open-weight model. Beam is a sparse Mixture-of-Experts (MoE) model with 501B total parameters and 23B active per token, built for coding, reasoning and agentic workloads. As per the Reflection AI team, Beam directly competes with larger open models like GLM 5.2 while using 3 to 4x less inference
Most production recommenders are cascades. Candidate generators feed a pre-ranker, which feeds a heavy ranker built on hundreds of engineered features. Yandex’s Sona Technical Report describes a different design. Sona is a generative AI model that brings candidate generation and ranking into a single system, replacing the multiple stages typically used in recommendation pipelines. Yandex
In April 2023, Alibaba Cloud demoed a chatbot whose name roughly means ‘truth from a thousand questions.’ Three and a half years later, its descendant ships open weights with 2.4 trillion parameters. This is the story of how Qwen got there, release by release. window.addEventListener(‘message’,function(e){if(e.data&&e.data.mtpQwenTl&&e.data.h){var f=document.getElementById(‘mtp-qwen-tl-frame’);if(f&&e.source===f.contentWindow){f.style.height=e.data.h+’px’;}}}); Chapter 1 — 2023: a thousand questions Alibaba moved
The Story of Qwen: Alibaba’s AI Models From 7B to 2.4T Read More »
Cantina Security, with Yeta Labs, has released apex-flash-1, an open-weights model trained specifically for vulnerability research. It is a reinforcement learning fine-tune of Z.ai’s GLM-5.3-Flash, released on Hugging Face under the MIT license. Is it deployable? Yes, the MIT weights serve on vLLM, SGLang or Transformers, but BF16 needs roughly 640 GB of GPU memory.
Anthropic, OpenAI and Google DeepMind shipped 4 frontier-class models within 30 days. Claude Fable 5.1 arrived on September 1. GPT-6 Astra followed on September 3. GPT-6.1 Sol and Gemini 4 Argon landed in the last days of September. We covered each launch on its own. This piece puts them side by side. The benchmark scores
Aleph Alpha has released Kolibri, an open-weight Mixture-of-Experts (MoE) language model built for German and English. Kolibri has 78.1B total parameters but activates only 3.46B, or 4.4%, per token. It accepts up to 1,048,576 tokens of context, lets users set reasoning effort per request, and ships under the Apache 2.0 license on Hugging Face. The
