Tencent hires WizardLM team, a Microsoft AI group with an odd history

Tech Crunch - May 13th, 2025

WizardLM, a Beijing-based AI research group formerly associated with Microsoft, has transitioned to Tencent's AI division, Hunyuan. This move was confirmed by Can Xu, a senior AI researcher, who announced on X that the team had departed Microsoft to integrate with Tencent's Hunyuan, a group known for developing advanced AI models like video and 3D object generators. The transition has already borne fruit with the release of the Hunyuan-TurboS 0416 model, claimed by Qingfeng Sun, a co-founder of WizardLM, to outperform Google's Gemma 3 series. The exact number of researchers involved in this transition and the timing of their departure from Microsoft remains unclear, pending comments from both companies.

WizardLM's history at Microsoft was marked by controversy when its AI model, WizardLM-2, was pulled from the web due to lack of toxicity testing, despite being competitive with OpenAI's GPT-4. This removal disrupted open-source projects on platforms like Hugging Face, sparking outcry from the community. Tencent's recent restructuring of its Hunyuan team, coupled with a significant increase in AI infrastructure spending, underscores its commitment to AI, contributing to an 8% year-over-year growth in Q1 2025. Tencent plans to invest around $12.49 billion in AI efforts this year, suggesting a robust future for WizardLM under its new banner.

Microsoft Tencent Ai Models Ai Development Hugging Face Wizard Lm Can Xu Hunyuan Qingfeng Sun Gemma 3

Story submitted by Fairstory

RATING

6.0

Moderately Fair

Read with skepticism

The article provides a timely and relevant overview of WizardLM's transition to Tencent, touching on significant developments in the AI sector. It presents factual claims that align with known information, but lacks comprehensive verification and diverse sourcing. The story is clear and accessible, yet could benefit from more detailed analysis and external perspectives to enhance its balance and credibility. While it addresses topics of public interest, its impact is limited by the absence of in-depth exploration of the broader implications. Overall, the article serves as a useful update for those following AI advancements, though it leaves room for further investigation and discussion.

Accuracy

The story presents several factual claims, such as the WizardLM team joining Tencent and the release of the Hunyuan-TurboS 0416 model. These claims are generally supported by available information, but there are areas requiring further verification. For example, the article mentions that Can Xu announced the team's move on X, which aligns with typical corporate announcements, yet lacks direct quotes or links to verify the statement. The claim that Hunyuan-TurboS 0416 outperforms Google's Gemma 3 series is bold and requires independent verification through comparative testing, which the article does not provide. Additionally, the story's reference to WizardLM's past incident with Microsoft over toxicity testing is consistent with known events, but it does not provide new insights or updates on the situation.

Balance

The article predominantly focuses on the perspective of WizardLM and their transition to Tencent, providing limited viewpoints from other stakeholders, such as Microsoft or independent analysts. By not including comments from Tencent or Microsoft, the article misses an opportunity to present a more balanced view. The narrative is primarily centered around the achievements and challenges of WizardLM, potentially overshadowing the broader implications for the AI industry or competitive landscape.

Clarity

The language and structure of the article are generally clear, with a logical flow that guides the reader through the narrative. However, the lack of detailed explanations for some technical claims, such as the performance of AI models, may hinder comprehension for readers unfamiliar with the subject matter. The tone is neutral, but the inclusion of more specific examples or data would enhance clarity and engagement.

Source quality

The sources cited in the article, such as statements from Can Xu and Qingfeng Sun, appear credible but are not directly quoted or linked, which affects their reliability. The story lacks a diverse range of sources, relying heavily on internal announcements from WizardLM members. Including perspectives from external experts or official comments from Tencent and Microsoft would enhance the credibility and depth of the reporting.

Transparency

The article does not provide sufficient context or methodology for some of its claims, such as the performance comparison between AI models. It also does not disclose any potential conflicts of interest or the basis for its assertions about Tencent's restructuring and AI investments. Greater transparency in these areas would improve the reader's understanding of the story's foundation and potential biases.