Alibaba releases Qwen 2.5-Max, an AI model that the company's cloud unit claims “outperforms” GPT-4o, DeepSeek-V3, and Llama-3.1-405B “almost across the board”
Chinese tech company Alibaba (9988.HK) on Wednesday released a new version of its Qwen 2.5 artificial intelligence model …
ReutersEduardo Baptista
Context & Ripple Effects
Alibaba had already expanded Qwen 2.5 while reporting broad corporate uptake; an outside evaluation in that coverage found strengths in language and creation but weaker knowledge and reasoning, providing context for the new model's broader performance claim. the earlier Qwen 2.5 rollout and its mixed external assessment
The release is an early step in a continuing Qwen cadence that later included a smaller open-source multimodal model for edge deployment and larger proprietary previews. It matters because Alibaba is using model releases to compete simultaneously on capability claims and deployment options.
First-order effects
Alibaba's cloud unit gains a new flagship model to position against GPT-4o, DeepSeek-V3 and Llama-3.1-405B; the reported advantage remains the company's claim rather than an independently established result.
Enterprise and developer buyers using or considering Qwen have another model option to evaluate against those named alternatives, with benchmark coverage likely central to the decision.
Second-order effects
The named rival model providers face added pressure to substantiate performance with comparable evaluations and to compete on dimensions beyond headline benchmark claims, including availability and deployment fit.
Alibaba's cloud sales effort can tie model selection more closely to its platform, making distribution and integration a more important competitive lever than a standalone model ranking.
Third-order effects
If repeated releases translate into credible performance and adoption, frontier-model competition will increasingly be shaped by a portfolio of proprietary and open deployments rather than a single best-model narrative.
As more suppliers make competing performance claims, buyers may gain leverage to demand clearer task-level evidence and better commercial terms; benchmark leadership alone may become less decisive.
The trend: Alibaba's Qwen releases are part of AI model competition shifting from isolated benchmark races toward platform-backed portfolios that combine capability, distribution and deployment choice.
The burst of DeepSeek V3 has attracted attention from the whole AI community to large-scale MoE models. Concurrently, we have been building Qwen2.5-Max, a large MoE LLM pretrained on massive data and post-trained with curated SFT and RLHF recipes. It achieves competitive [image]
Today marks the Chinese New Year, and while fireworks light up the sky outside, here I am, sitting in front of my computer, writing this post. We've finally released Qwen2.5-Max, an MoE model on par with Deepseek-V3, now available on Qwen Chat and via API. That's right—we're
Qwen2.5-Max is here. Looks good at benchmarks and I hope you guys can give it a try and see how you feel about this new model! Qwen Chat: https://chat.qwenlm.ai/ (choose Qwen2.5-Max for the model) API is available through Alibaba Cloud service.| Happy new year!
Can someone ask China to take a break? 🤯 Qwen just dropped another model, Qwen2.5-Max, which outperforms DeepSeek V3 You can try it for free on Qwen Chat and also available on API. More details below 👇 [video]
Results of base language models. We are confident in the quality of our base models and we expect the next version of Qwen will be much better with our improved post-training methods. [image]
Actually compared with Qwen2.5-Max with more attention, I love Qwen2.5-VL as much. We guys suffered quite a bit in either data and training, we have tried the best to strike a balance between benchmarks and human preferences. I think this time you will find more surprises in the
Qwen2.5-Max just one shotted this prompt: write a script for three bouncing yellow balls within a sphere, make sure to handle collision detection properly. make the sphere slowly rotate. make sure balls stays within the sphere. implement it in p5.js developers can start using [vi…