Tag
Meta released Muse Spark Open, an 816B-parameter (42B active) open-weight multimodal agent model with a 1M-token context, but the author argues its marketing is misleading since the hardware required to run it locally costs tens of thousands of dollars, making API access the realistic option for most users.
BAAI releases AREX-2, a 27B long-horizon self-improving agent model based on Qwen3.8 that learns to propose, measure, reflect, and revise solutions over multiple test-time rounds, achieving strong results on coding, ML engineering, and deep research benchmarks with only 27B parameters.
This article details the local deployment guide for the Qwen3.8 27B model, covering two routes for Mac and Nvidia graphics cards, and provides real-world performance data to help users run this model on consumer-grade hardware.
Owl Alpha, a rapidly growing agent model on OpenRouter, is reportedly a rebranded version of Meituan LongCat-2.0-Preview, a massive 1.6T-parameter MoE model with 48B active parameters and 1M-token context. It has achieved notable usage rankings and growth rates.
Julien C asks about open-source release of Alibaba's Qwen3.7-Plus, a multimodal agent model unifying vision and language.
Poolside is hosting a 2-day model research hackathon in London to push an open-weight agent model further using RL and fine-tuning on Laguna XS.2, with partners including NVIDIA, Prime Intellect, and Hugging Face, and a prize of an NVIDIA DGX Spark.
UI-TARS-2 is a native GUI-centered agent model that addresses data scalability, multi-turn RL, and environment stability challenges, achieving state-of-the-art results on GUI benchmarks (88.2 on Online-Mind2Web, 47.5 on OSWorld, 50.6 on WindowsAgentArena,73.3 on AndroidWorld) and outperforming Claude and OpenAI agents.