阿里发布新一代千问旗舰模型Qwen3.7-Max

This article summarizes Alibaba's newly launched Qwen3.7-Max, the highest-performing domestic large language model in China, whose capabilities will soon become available to general users. Key takeaways are as follows:

1. Performance: The model ranks first among all Chinese-developed large models in global blind benchmark tests, with performance close to top overseas models including GPT and Claude. It outperforms other Chinese models on authoritative evaluations across coding, reasoning, multilingual understanding and agent intelligence, and surpasses leading overseas model Claude-Opus 4.6 on some metrics.

2. Core capabilities: Designed specifically for agent applications, the model can independently complete complex long-horizon tasks lasting up to 35 hours. It can optimize inference kernels from scratch on brand-new chips and achieve a 10x speedup in inference, as well as enable office automation. Complex projects that previously took professional teams 1 to 2 weeks can now be delivered within a few hours.

3. Access: Qwen3.7-Max's API will soon be launched on Alibaba Cloud Bailian. Multiple versions will be released in the future to support multimodal requirements, and both general users and developers will be able to access the model via API calls.

Alibaba's release of its new generation Qwen large model opens up new opportunities for AI-powered upgrading for brands. Below is a summary of key takeaways tailored to brands' core concerns including marketing, R&D and consumer trends:

1. Ready-to-use technology foundation: Qwen3.7-Max is currently the highest-performing domestic large model in China, with leading performance in coding, reasoning and agent intelligence. Its API will open for calls soon, allowing brands to develop custom AI applications on top of this foundation to support their own digital transformation.

2. Clear efficiency gains: With powerful workflow automation capabilities, the model drastically cuts delivery cycles for complex projects, helping brands cut costs and boost efficiency. It can improve performance across content marketing, user operations, and product R&D.

3. Clear future outlook: The AI industry has now entered the agentic era. The Qwen 3.7 series will launch versions supporting multimodal inference and visual understanding to cover needs across a wide range of scenarios. Brands can make early布局 of AI empowerment to capture growth opportunities brought by this new round of technological upgrading.

Alibaba's release of the Qwen3.7-Max large model brings clear AI-powered growth opportunities for all types of sellers. Below is a summary tailored to sellers' core concerns:

1. Access top-tier capabilities at low cost: Qwen3.7-Max's API will soon be available on Alibaba Cloud Bailian. Sellers do not need to invest heavily in R&D to access the capabilities of China's top large model, which can be used for a wide range of routine tasks including marketing copy generation, customer service, and data analysis to improve operational efficiency.

2. Enable fully automated operations: With powerful agent and workflow automation capabilities, the model can complete complex tasks independently. Sellers can build automated operation systems based on the model to replace manual labor for large volumes of repetitive work and cut labor costs. Complex projects that previously took one to two weeks can now be delivered in just a few hours.

3. Clear ecosystem opportunities: Alibaba Cloud has completed a full system upgrade to adapt to the agentic era, and the Qwen series has delivered three stable iterations in three months, showing extremely fast technological upgrading. Sellers can leverage the Alibaba Cloud ecosystem to capture new growth opportunities from AI empowerment. No explicit public risks have been announced so far.

Alibaba's new generation Qwen large model brings multiple insights for factories pursuing digital transformation and new business opportunities. Key takeaways are as follows:

1. Lower the barrier to digital transformation: With powerful coding and complex task processing capabilities, the model can independently complete professional-grade engineering optimization work. Development projects that previously took professional engineering teams one to two weeks can be delivered by the model-driven agent in just a few hours, helping factories cut the development cost and lead time for custom digital systems.

2. Adapt to multi-scenario requirements: Supporting multi-agent collaboration and enabling end-to-end workflow automation, the model fits automation needs across multiple factory scenarios including production management, supply chain coordination and office operations, helping factories improve overall operational efficiency.

3. Leverage a mature ecosystem: Alibaba Cloud has built a complete "chip-cloud-model-inference" technology stack. Factories do not need to build large model capabilities from scratch, and can advance transformation based on this mature foundation. Future multimodal versions will also support more manufacturing scenarios including product design and industrial inspection.

This release reveals the latest development trends of the AI industry and brings new opportunities for AI service providers. Below is a summary tailored to service providers' core concerns about industry trends, pain points and solutions:

1. Clear industry development trend: The AI industry has now entered the agentic era, where coding capability and long-horizon complex task processing have become core competitiveness of large models. Chinese large models are iterating at an extremely fast pace: the Qwen series has delivered three iterations in three months, and Qwen3.7-Max has reached performance close to top overseas models, even outperforming them on some metrics, accelerating industrial adoption.

2. Address unmet customer needs: Many enterprise clients need large model foundations that can support long-cycle complex tasks, while existing models cannot handle dozens of hours of independent iteration work. Qwen3.7-Max fills this gap and can support the development of all types of enterprise-grade AI applications.

3. Cut costs and boost efficiency by building on the foundation: Qwen3.7-Max will soon launch its API on Alibaba Cloud Bailian, and is compatible with all mainstream agent frameworks. Multimodal versions will follow. Service providers can develop industry solutions directly on this foundation, drastically cutting their own R&D costs.

Alibaba Cloud's latest upgrade and new model release provides a lot of valuable insights for all types of platform operators. Below is a summary tailored to platforms' core concerns about demand, operations and industry trends:

1. Clarifies market demand direction: The core market demand for AI platforms has now shifted to supporting agent applications. Clients require large model foundations with stronger performance that can handle long-cycle complex tasks. Platforms need to follow this technological trend, adjust their product layout, and adapt to the needs of the agentic era.

2. Actionable benchmark from Alibaba Cloud's strategy: Alibaba Cloud has completed a full upgrade for the agentic era, building a complete "chip-cloud-model-inference" technology stack, opening large model APIs via Alibaba Cloud Bailian, and offering a tiered product line of models with different specifications to cover the needs of different customer groups. This commercial and operational strategy can serve as a reference for all types of platforms.

3. Highlights risks to avoid: Large model technology is currently iterating extremely quickly: the Qwen series has delivered three stable iterations in just three months, continuously pushing up the performance ceiling. Platforms that fail to keep up with technological upgrades will quickly lose users. They must continue investing in R&D, follow the latest technology trends, and open up their capabilities to attract developers to enrich their ecosystem.

This release reveals the latest industry trends of the large model sector and brings multiple insights for AI-related research. Below is a summary tailored to researchers' core concerns about industry trends, business models and new technologies:

1. Clear new industry trends: Chinese large models are iterating faster than any other region in the world, with extremely rapid performance gains. Qwen3.7-Max has outperformed leading overseas model Claude-Opus 4.6 on multiple core benchmarks, ranking as the top-performing Chinese-developed large model, with performance close to the world's strongest overseas models including GPT and Gemini. This confirms that Chinese large models have now entered the global first tier.

2. Opens new directions for technical research: Qwen3.7-Max can complete 35 hours of long-horizon independent task iteration, optimize inference kernels from scratch on brand-new chips to achieve 10x inference speedup, and even proactively redesign its own architecture. This proves that large models are capable of long-cycle autonomous evolution, opening up entirely new research directions for large model autonomous evolution.

3. Provides a new sample for business model research: Alibaba Cloud has built a complete "chip-cloud-model-inference" technology stack, and pursues tiered commercialization via open API access, with different model specifications serving different demands. This go-to-market model provides a new research sample for large model commercialization in the agentic era.