How to Use Qwen and Alibaba’s AI to Destroy ChatGPT

In recent years, the generative artificial intelligence sector has experienced an unprecedented surge, long dominated by a few Western players. But 2026 marks a turning point. Alibaba has significantly accelerated its AI efforts with the release of Qwen 3.5, a profoundly revamped version of its intelligent assistant that no longer simply "answers well," but aims to become a truly autonomous agent, integrated into the everyday digital economy.

In recent days, Alibaba officially launched Qwen as a direct evolution of Tongyi, its artificial intelligence chatbot already available on the iPhone App Store and the Android Play Store. The Chinese company presented this new product, calling it the most powerful AI assistant on the market—a statement that today, with the arrival of version 3.5, takes on much more concrete weight than in the past.

The stated goal remains to challenge giants like ChatGPT, but the strategy has changed: not just high performance, but also architectural efficiency, cost reduction, and a strong focus on open source. Qwen 3.5 was born in a context of fierce competition, also involving other Chinese giants like ByteDance and MiniMax, and at a delicate geopolitical moment for Alibaba, which has come under scrutiny due to potential tensions with the United States.

But how does Qwen really work today? What are its distinctive capabilities compared to the past? And above all: is it already better than ChatGPT in some specific areas? Here’s everything you need to know.

What is Qwen?

Qwen is Alibaba’s proprietary artificial intelligence chatbot and today represents much more than a simple conversational assistant. From the earliest versions, the company has confirmed and strengthened an open source, or rather open-weight, approach, making the model’s trained weights available to allow developers and companies to customize it, deploy it locally, and integrate it into complex environments.

With the arrival of Qwen 3.5, this approach takes on new strategic importance. The flagship model, Qwen3.5-397B-A17B, has 397 billion total parameters, but activates only 17 billion of them for each inference thanks to a hybrid architecture based on mixture-of-experts and linear attention. In practice, a "frontier" scale model with operating costs comparable to much smaller systems.

This choice allows Alibaba to offer different versions: from lightweight models, suitable for local execution on workstations or corporate servers, to extremely powerful cloud variants, capable of managing very long contexts and complex workflows.

It’s no coincidence that Qwen is now described as a comprehensive platform, integrated into the Alibaba ecosystem, from e-commerce to logistics, from payments to consumer services.

How Qwen works with the arrival of the 3.5 model

One of Qwen’s great advantages remains its multimodal nature. With version 3.5, this feature becomes native: the model is able to process text, images, and videos within a single reasoning process, managing content up to two hours long or extensive technical documentation in a single session.

Computational thinking has been defined as hybrid. In practice, Qwen alternates between a deep thinking model and a quick response, dynamically choosing the best approach based on the complexity of the prompt.

This behavior is further refined thanks to the so-called thinking budget, which allows for the allocation of more or fewer reasoning tokens for each request, balancing speed and depth.

From a technical standpoint, Qwen 3.5 introduces innovations such as multi-token prediction, which reduces latency by generating multiple tokens per cycle, and optimized FP8 computation pipelines, capable of reducing energy consumption and hardware costs without compromising stability.

Regarding information, the real-time web search tool remains central, designed to provide always-updated answers. On the linguistic front, support now extends to over 200 languages and dialects, making Qwen one of the most inclusive models available.

How to use Qwen

But in practical terms, how do you use Qwen today and why does version 3.5 mark a clear departure from previous versions? Alibaba has worked to make access to Qwen extremely flexible, adapting it to both consumer users and complex professional and business contexts.

On the consumer front, Qwen is accessible via Qwen Chat, available via browser and mobile app. The interface is designed to be immediate, but behind this simplicity lies a much more sophisticated structure than in the past. With Qwen 3.5, users can choose between three operating modes:

• Auto, which enables adaptive thinking and automatic use of tools;
• Thinking, designed to activate deep reasoning and multi-step planning;
• Fast, optimized for rapid response and reduced latency.

In the prompt entry bar, as with major competitors, you can upload documents, images, videos, and audio files. However, the real difference lies in the handling of long contexts. The hosted version of Qwen 3.5 Plus reaches a context window of 1 million tokens, making it possible to analyze entire codebases, extensive business reports, or very long video content in a single session.

For developers and enterprises, Qwen 3.5 can be integrated via API through Alibaba Cloud Model Studio. This brings advanced features such as extended reasoning, real-time search tools, and the management of autonomous agents capable of executing complex workflows. Unlike many proprietary models, Qwen also offers the possibility of local deployment thanks to its open-weight distribution, allowing companies to maintain full control over their data and infrastructure, a particularly important aspect in regulated industries.

The ecosystem is complemented by support for tools such as Ollama and LM Studio, which enable local execution and facilitate technical audits, fine-tuning of proprietary data, and greater privacy protection. This is not just a technical choice, but also a strategic one: in a European context increasingly focused on digital sovereignty and the AI Act, the ability to run an advanced model under controlled jurisdiction represents a concrete competitive advantage.

At the application level, Qwen demonstrates a strong industrial focus. Customer service is one of the most mature areas, with models trained to manage chatbots integrated into e-commerce platforms. Thanks to Qwen-Audio, voice interactions are also available, while in the healthcare and financial sectors, Qwen is already being used for document analysis, diagnostic support, automated report creation, and data analysis.

The Differences Between Qwen and ChatGPT

The key question remains the same: Why choose Qwen over ChatGPT? At a superficial glance, the differences may seem minimal. Both are multimodal models, capable of understanding text, images, and complex content, and both offer advanced conversational interfaces. It’s no coincidence that many have dismissed Qwen as the "Chinese copy of ChatGPT".

In reality, this definition ignores the most important difference: the architecture and development philosophy. Qwen is an open-weight model, with publicly available trained weights, while ChatGPT is a closed, proprietary system, accessible only through interfaces and APIs controlled by OpenAI. This means that Qwen can be downloaded, analyzed, modified, and adapted to specific needs, while ChatGPT cannot.

From a technical standpoint, Qwen 3.5 focuses on architectural efficiency rather than brute-force scaling. The Qwen3.5-397B-A17B model, despite having 397 billion total parameters, activates only 17 billion of them for inference thanks to sparse mixture-of-experts and linear attention. The result is a system that, according to declared benchmarks, achieves performance comparable to much larger proprietary models, but with significantly lower computational costs, energy consumption, and latency.

ChatGPT, for its part, remains more widespread and immediate, especially for general users. Its interface is extremely fluid and hides much of the complexity of reasoning, offering concise and well-structured responses. This makes it ideal for everyday use and for those seeking rapid support without the need for customization.

Qwen, on the other hand, is firmly focused on the enterprise world. Its open-weighted orientation enables in-depth audits, fine-tuning on proprietary data, and integration into complex industrial pipelines. In an increasingly stringent regulatory environment, this possibility of direct control becomes a discriminating factor, especially in Europe.

The economic aspect must also be considered. Alibaba is openly engaging in a price war, reducing the cost of tokens and making AI accessible to even small and medium-sized businesses. This approach, combined with the ability to execute locally, positions Qwen as a sustainable solution in the long term.

Ultimately, there is no clear winner. ChatGPT remains the most popular choice for the standard user. Qwen 3.5, on the other hand, is aimed at those seeking power, control, and deep integration, accepting a slightly steeper learning curve in exchange for greater freedom.

Original article published on Money.it Italy. Original title: Come usare Qwen e come funziona l’AI di Alibaba che punta a distruggere ChatGPT