According to TechCrunch, the artificial intelligence research team WizardLM, which was headquartered in Beijing and formerly affiliated with Microsoft, has reportedly joined Tencent AI Lab's "Hunyuan" team as a whole. Several core researchers announced this move on social platform X (formerly Twitter), signaling that Tencent is further strengthening its offensive in the large model field.
In his Tuesday post, Can Xu, a senior researcher at Microsoft who once led multiple projects of WizardLM, stated that he and his team have left Microsoft and officially joined Tencent's Hunyuan team. This team has released groundbreaking models in multiple directions such as video generation and 3D content creation in recent years.
What draws more attention is that WizardLM recently publicly released a model named Hunyuan-TurboS0416, which is the first time that the product name has appeared with the word "Hunyuan", indicating that the team affiliation has deeply integrated with Tencent. According to Sun Qingfeng, who claims to be one of the co-founders of WizardLM, the model has surpassed Google's latest open-source large model Gemma3 series in multiple benchmark tests, and is considered a key achievement for Tencent in the open-source competition.
The aftereffects of Microsoft's "accidental deletion" incident are still lingering
Since its establishment, WizardLM has been controversial. In April 2024, the team released its self-developed large model WizardLM-2 and claimed that its capabilities were close to GPT-4, but it was taken down by Microsoft the next day due to "incomplete toxicity testing". Although the team urgently supplemented the tests and promised to re-release, the original model had already been rapidly spread and re-uploaded by the community, making it difficult for Microsoft to control.
This event also sparked strong dissatisfaction in the open-source circle. Clément Delangue, CEO of Hugging Face, publicly criticized Microsoft's takedown behavior for "destroying dozens of open-source projects," emphasizing that the monthly download volume of the WizardLM model once exceeded 100,000 times, having a huge impact on the ecosystem.
Reorganization of the Hunyuan team and upgrade of Tencent's AI strategy
At the time when the WizardLM team joins, Tencent is accelerating the layout of the large model ecosystem. It is reported that Tencent has reorganized the Hunyuan large model team into two independent departments and expanded infrastructure investment.
In the just-released Q1 2025 financial report, Tencent explicitly attributed 8% of revenue growth year-over-year to the AI strategy upgrade and announced that it will invest 90 billion RMB (approximately $12.49 billion) in capital expenditures for the year, most of which will be used for AI capability construction.
Tencent's AI team expansion, especially the absorption of elite teams like WizardLM with experience in large model development and open source, indicates that it no longer wants to remain a follower and is attempting to seize more dominance in the global AI model competition.
It remains unclear whether the entire WizardLM team has joined Tencent, and their official resignation time from Microsoft is also unknown. Neither Microsoft nor Tencent has made a public statement about this change.