Article

Explore our curated collection of AI articles covering artificial intelligence, machine learning, AI tools and more

284 Article•Page 1 / 32

P1-VL: Open-Source Vision-Language Model

P1-VL, an open-source vision-language model designed for physical reasoning, is built upon Qwen3-VL-235B-A22B-Thinking and optimized through multi-stage reinforcement learning on curated physics competition data.

Open source

MOSSAudioTokenizer: Discrete Audio Tokenizer

MOSSAudioTokenizer, a discrete audio tokenizer based on the Cat (Causal Audio Tokenizer with Transformer) architecture. It scales up to 1.6 billion parameters and serves as a unified discrete interface that enables both lossless quality reconstruction and high-level semantic alignment.

Open source

MiniMax-M2.5, the latest open-source model from MiniMax

MiniMax-M2.5, the latest open-source model from MiniMax, achieves state-of-the-art performance in coding, agent tool usage and search, office work, and a range of other economically valuable tasks.

Open source

DeepGen 1.0: A Lightweight Unified Multimodal Model

DeepGen 1.0 is a lightweight, unified multimodal model that supports general image generation, general image editing, reasoning-based image generation, reasoning-based image editing, and text rendering.

Open source

Xiaomi-Robotics-0: Xiaomi's Open-Source Vision-Language-Action Model

Xiaomi-Robotics-0, an open-source vision-language-action model from Xiaomi with 4.7 billion parameters, is designed for high-performance robotic reasoning and seamless real-time execution.

Open source

ArcFlow: Unlocking Two-Step Text-to-Image Generation via High-Precision Nonlinear Flow Distillation

By leveraging high-precision nonlinear flow distillation, we enable two-step text-to-image generation within a multi-step distillation framework that explicitly employs nonlinear flow trajectories to approximate the pre-trained teacher's trajectory.

Open source

Ming-flash-omni 2.0: Open-source Multimodal Model

Ming-flash-omni 2.0 is an open-source multimodal model that supports visual encyclopedic knowledge, immersive speech synthesis, and high-dynamic-range image generation and editing.

Open source

UI-Venus 1.5: End-to-End GUI Agent

UI-Venus 1.5, an end-to-end GUI agent designed for robust real-world applications. This model family includes two dense variants (2B/8B) and one MoE variant (30B-A3B) to meet diverse downstream scenarios.