
Alibaba Unveils Qwen3.6-Plus Agent Model
Alibaba Cloud launches Qwen3.6-Plus with 1 million token context window.
The cloud division of Chinese tech giant Alibaba has launched Qwen3.6-Plus. The model offers a default context window of 1 million tokens, enhanced agent capabilities for programming, and more accurate multimodal perception and reasoning.
The chatbot is available via the Alibaba Cloud Model Studio API and is compatible with third-party coding assistants, including OpenClaw, Claude Code, and Qwen Code.
Performance and Multimodality
Qwen3.6-Plus integrates reasoning, memory, and execution. In code correction tests, the model closely approached market leaders, and in task automation and complex terminal operations, it outperformed competitors.

The new version demonstrates some of the best results in long-term planning and the use of external tools. Qwen3.6-Plus also sets records in assessments of STEM reasoning, extracting information from ultra-long contexts, and multilingual environments.
Alibaba continues to expand the product’s capabilities, shifting focus from basic machine vision to complex multimodal analysis and autonomous AI agent operation.
The updated model better understands the structure of complex documents, writes code, and analyzes videos and physical world objects. Qwen3.6-Plus can identify connections between information arrays and independently determine the optimal algorithm for solving a given task.

Based on screenshots of interfaces, product prototypes, design layouts, or multimodal instructions, the model generates frontend pages.
Ecosystem and Integrations
The model is compatible with several popular tools:
- Qwen Code — an open-source AI agent for terminal work, specially adapted for this series. It helps analyze complex code bases, automate routine tasks, and speed up development;
- Claude Code — the Alibaba Cloud API supports the Anthropic API protocol. This allows the use of Qwen3.6-Plus with tools like Claude Code for an enhanced programming experience.
The new version of the API includes a preserve_thinking function (disabled by default). It retains the intermediate thought process of the neural network from all previous dialogue stages. Developers recommend using it in agent scenarios: continuous reasoning context makes responses more accurate and often reduces overall token consumption.
Alibaba Cloud Model Studio supports protocols compatible with OpenAI and Anthropic specifications.
Back in February, the previous version of the AI model — Qwen-3.5 — was released. At that time, the tool first gained native multimodality.
Рассылки ForkLog: держите руку на пульсе биткоин-индустрии!