Just as DeepSeek was making headlines, another Chinese AI model has emerged to challenge its position. Alibaba, the renowned e-commerce giant, has introduced a new iteration of its Qwen AI model series, and it’s already turning heads—surpassing both DeepSeek and ChatGPT in several critical aspects.

Introducing Qwen Max

Qwen Max is the flagship model in Alibaba’s Qwen AI family, representing the most advanced and capable version to date. The Qwen series currently includes the following models:

• Qwen2.5-Plus
• Qwen2.5-Max
• Qwen2.5-VL-72B-Instruct
• Qwen2.5-14B-Instruct-1M
• QVQ-72B-Preview
• QwQ-32B-Preview
• Qwen2.5-Coder-32B-Instruct
• Qwen2.5-Turbo
• Qwen2.5-72B-Instruct

These models are accessible for free after creating an account via email, Google, or GitHub. Additionally, the Qwen AI models are open-source, available on platforms like GitHub and HuggingFace. Users can even install them locally on their devices (depending on hardware capabilities) to operate the AI offline.

Key Features of Qwen Max

Qwen2.5-Max is a 72-billion parameter Mixture-of-Experts (MoE) model, supporting 29 languages and trained on over 20 trillion tokens. It can handle up to 128,000 tokens in a single conversation, making it ideal for processing lengthy documents or complex datasets. The model also excels at working with structured data formats, such as tables, CSVs, and JSON files.

Each Qwen model is tailored for specific tasks. For instance, Qwen2.5-Coder-32B-Instruct is optimized for coding, while QwQ-32B-Preview focuses on reasoning and problem-solving. While not every model is a jack-of-all-trades, most are capable of handling text prompts, image generation, and even video creation.

One standout feature is the ability to combine two models for enhanced performance. In practice, pairing models like Qwen2.5-Max with Qwen2.5-Coder-32B-Instruct can yield better results, such as generating code with fewer errors and requiring fewer prompts.

Why Qwen Stands Out

Alibaba’s Qwen series is not just another addition to the AI landscape—it’s a significant leap forward. With its multilingual support, massive token capacity, and specialized models, Qwen is positioning itself as a strong competitor to established AI tools like DeepSeek and ChatGPT. Whether you’re a developer, data analyst, or creative professional, Qwen offers a versatile and powerful solution for a wide range of tasks.

Accessing Qwen and Its Performance
The official platform to access Qwen’s AI models is through its dedicated website. Here, users can input prompts and work with text, as well as generate images and videos in various aspect ratios. While the platform also plans to introduce a Web Search feature, it has not yet been launched.

One notable drawback is that Qwen can sometimes be slow to process prompts, especially during the initial interaction. At first, I even wondered if the website was malfunctioning. The first prompt in a conversation often takes around 30 seconds to generate a response, though subsequent replies are noticeably faster.

When it comes to generating images and videos, Qwen performs quicker than expected. While the quality and realism may not be top-tier, the tool is handy for creating quick visuals in a pinch. That said, users should anticipate some random artifacts in the generated media.

Connectivity and Server Issues
The delayed response times might be attributed to server load, a common issue for many AI chatbots at launch, including DeepSeek and ChatGPT. During testing, I occasionally encountered connectivity errors due to an overload of requests in the queue.

How Does Qwen Compare to DeepSeek?
Technically, Qwen outperforms DeepSeek in several areas. Its interactions feel more natural, and it operates slightly faster. However, if you look beyond benchmark results, the differences between the two models become less apparent.

Qwen’s standout advantage is its alignment with human preferences, making it easier to input complex prompts and receive accurate responses without extensive fine-tuning. Even simple prompts can yield detailed and informative answers. In contrast, DeepSeek often requires multiple attempts and careful prompt engineering to achieve the desired outcome.

In terms of general knowledge and factual accuracy, both models are comparable, though Qwen has a slight edge in maintaining factual consistency.

Cost Comparison
Where DeepSeek clearly takes the lead is in cost efficiency. DeepSeek charges $0.25permilliontokens,whileQwencosts$0.38. Despite this, both models are significantly more affordable than GPT-4o ($5permilliontokens)andClaude3.5($3 per million tokens).

In summary, Qwen offers a robust and user-friendly AI experience, particularly for those seeking detailed and accurate responses. However, its higher cost and occasional connectivity issues may be considerations for users comparing it to alternatives like DeepSeek.

Benchmark Comparisons
As previously highlighted, Qwen consistently outperforms DeepSeek across a wide range of benchmarks, showcasing its superior capabilities in understanding and aligning with human values. Below is a detailed comparison of their performance across key benchmarks:

The data clearly demonstrates Qwen’s edge in areas such as reasoning, general knowledge, and coding, though the differences are relatively modest. Beyond benchmarks, Qwen’s ability to align more closely with human preferences gives it a distinct advantage, making it easier to use for complex tasks without extensive fine-tuning.

Setting New Standards in AI Development

Both Qwen and DeepSeek represent significant advancements in AI technology, setting new benchmarks for performance and innovation. However, their rise has also raised concerns about security and privacy. For instance, DeepSeek has already experienced a data breach, highlighting potential vulnerabilities.

Despite these concerns, the performance of these Chinese AI models has undoubtedly shaken the global AI landscape. They have proven to be formidable competitors to Western counterparts, offering superior performance in many areas. As the AI race intensifies, Qwen and DeepSeek are pushing the boundaries of what’s possible, leaving the tech world eagerly watching their next moves.