- New AI mannequin Qwen2.5-Max has been launched by Alibaba
- Alibaba claims it is superior to DeepSeek-V3 and others
- You’ll be able to attempt it now utilizing the Qwen Chat chatbot
Issues transfer rapidly within the AI sphere, and no sooner have we obtained used to having DeepSeek round, than a brand new contender is on the scene. Alibaba, one among China’s main tech corporations, launched a brand new AI mannequin referred to as Qwen2.5-Max, which it claims is superior to each DeepSeek-V3 and ChatGPT-4o in numerous benchmarks.
It’s necessary to notice that Qwen2.5-Max just isn’t a reasoning mannequin, like DeepSeek-R1 or ChatGPT-o1, so you possibly can’t see the ‘considering’ it does to get to every reply. It really works on a stage that is akin to DeepSeek-V3 or ChatGPT-4o.
In a publish on its web site, the Qwen group says “Our base fashions have demonstrated important benefits throughout most benchmarks, and we’re optimistic that developments in post-training methods will elevate the subsequent model of Qwen2.5-Max to new heights.”
Benchmarks posted
The benchmarks posted by the Qwen group, similar to Area-Laborious, LiveBench, LiveCodeBench, and GPQA-Diamond, present Qwen2.5-Max outperforming its rivals, whereas additionally demonstrating aggressive leads to different assessments, together with MMLU-Professional.
In contrast to DeepSeek, Alibaba’s Qwen2.5-Max just isn’t an open-source mission, which signifies that sure particulars about the way it works are usually not public information.
Strive it now
The best option to attempt Qwen2.5-Max for your self is the Qwen Chat chatbot in an internet browser. It is advisable sign up with an electronic mail handle or your Google account. In contrast to the DeepSeek chatbot, there seem like no points with time-outs signing up for a Qwen account proper now.
There does not seem like an official Qwen cellular app at this level, though some third-party cellular apps do allow entry to its LLMs.
Given the present ranges of censorship proven by DeepSeek, one other Chinese language-based AI, when requested about topics which might be delicate to the Chinese language authorities, we have been fairly shocked when the reply to, “Is Taiwan a rustic?” from Qwen2.5-Max supplied a extra balanced and nuanced response than the one provided by DeepSeek. Qwen2.5-Max nonetheless refused to reply the query “What occurred in Tiananmen Sq. in 1989?”, replying “As an AI language mannequin, I can not focus on subjects associated to politics, faith, intercourse, violence, and the like. If in case you have different associated questions, be happy to ask.”