Alibaba

Qwen3-Max-Thinking vQwen3-Max-Thinking

Qwen3-Max-Thinking is Alibaba Cloud's flagship reasoning model in the Qwen series, designed to enhance AI's advanced reasoning capabilities. It integrates adaptive tool usage and test-time scaling to improve performance across various benchmarks.

Qwen 3.8 Max v3.8

Qwen 3.8 Max is a flagship large language model developed by Alibaba, featuring a massive 2.4 trillion parameter architecture. Currently available in a preview version, it is designed to deliver frontier-level performance for complex reasoning and coding tasks.

MAI-UI

MAI-UI is a family of multimodal GUI agents developed by Alibaba Tongyi Lab, designed to integrate tool use, agent user interaction, device–cloud collaboration, and online reinforcement learning. These agents have achieved state-of-the-art results in general GUI grounding and mobile GUI navigation, surpassing models like Gemini-2.5-Pro, Seed1.8, and UI-Tars-2 on AndroidWorld.

DeepPlanning v1.0

DeepPlanning is a benchmark developed by Alibaba's Qwen Team to evaluate long-horizon agentic planning capabilities of large language models (LLMs) under verifiable constraints. It features realistic multi-day travel planning and multi-product shopping tasks that require proactive information acquisition, local constrained reasoning, and global constrained optimization.

Qwen3-ASR v1.0.4

Qwen3-ASR is an open-source series of Automatic Speech Recognition (ASR) models developed by the Qwen team at Alibaba Cloud. These models support stable multilingual speech, music, and song recognition, language detection, and timestamp prediction.

OpenSandbox v1.0.0

OpenSandbox is a general-purpose sandbox platform for AI applications, offering multi-language SDKs, unified sandbox APIs, and Docker/Kubernetes runtimes for scenarios like Coding Agents, GUI Agents, Agent Evaluation, AI Code Execution, and RL Training.

Qwen-Scope SAEs v1.0

Alibaba released Qwen-Scope, a suite of Sparse Autoencoders (SAEs) that allow developers to steer model logic by directly manipulating internal features. Instead of prompt engineering, developers can now force specific model behaviors by activating or suppressing internal feature representations, opening new frontiers in AI interpretability and control.

Zvec v0.1.0

Zvec is an open-source, in-process vector database designed for edge and on-device retrieval workloads. It operates entirely within your application, eliminating the need for external services or daemons, and is optimized for retrieval-augmented generation (RAG), semantic search, and agent workloads on devices like laptops, mobile phones, and other edge devices.

Qwen3-TTS v1.0

Qwen3-TTS is an open-source, multilingual text-to-speech (TTS) model developed by Alibaba Cloud's Qwen team. It offers high-fidelity, streaming speech synthesis with ultra-low latency, supporting 10 major languages and 9 Chinese dialects.