New DeepSeek AI rival claims to be more powerful than both V3 and ChatGPT-4o – meet Qwen2.5-Max

(Image credit: Qwen/Alibaba)

New AI model Qwen2.5-Max has been released by Alibaba
Alibaba claims it's superior to DeepSeek-V3 and others
You can try it now using the Qwen Chat chatbot

Things move quickly in the AI sphere, and no sooner have we got used to having DeepSeek around, than a new contender is on the scene. Alibaba, one of China’s leading tech companies, released a new AI model called Qwen2.5-Max, which it claims is superior to both DeepSeek-V3 and ChatGPT-4o in various benchmarks.

It’s important to note that Qwen2.5-Max is not a reasoning model, like DeepSeek-R1 or ChatGPT-o1, so you can’t see the ‘thinking’ it does to get to each answer. It works on a level that's comparable to DeepSeek-V3 or ChatGPT-4o.

In a post on its website, the Qwen team says “Our base models have demonstrated significant advantages across most benchmarks, and we are optimistic that advancements in post-training techniques will elevate the next version of Qwen2.5-Max to new heights.”

Benchmarks posted

Qwen2.5-Max benchmarks — Benchmarks comparing the performance of the instruct models for Qwen2.5-Max vs rivals like Llama-3.1, DeepSeek-V3 and ChatGPT-4o. (Image credit: Qwen/Alibaba)

The benchmarks posted by the Qwen team, such as Arena-Hard, LiveBench, LiveCodeBench, and GPQA-Diamond, show Qwen2.5-Max outperforming its rivals, while also demonstrating competitive results in other assessments, including MMLU-Pro.

Unlike DeepSeek, Alibaba’s Qwen2.5-Max is not an open-source project, which means that certain details about how it works are not public knowledge.

Try it now

The easiest way to try Qwen2.5-Max for yourself is the Qwen Chat chatbot in a web browser. You need to sign in with an email address or your Google account. Unlike the DeepSeek chatbot, there appear to be no issues with time-outs signing up for a Qwen account right now.

There doesn't appear to be an official Qwen mobile app at this point, although some third-party mobile apps do enable access to its LLMs.

Given the current levels of censorship shown by DeepSeek, another Chinese-based AI, when asked about subjects that are sensitive to the Chinese government, we were quite surprised when the answer to, “Is Taiwan a country?” from Qwen2.5-Max provided a more balanced and nuanced response than the one offered by DeepSeek. Qwen2.5-Max however refused to answer the question “What happened in Tiananmen Square in 1989?”, replying “As an AI language model, I cannot discuss topics related to politics, religion, sex, violence, and the like. If you have other related questions, feel free to ask."