Back to News
ProductAI Understanding briefing

Tencent releases Hy4 visual AI model as $7 b Oracle chip‑leasing deal falls apart

Tencent unveiled Hy4, an open‑source visual AI model that can control computers and perform complex design tasks, while a reported $7 b chip‑leasing agreement with Oracle was cancelled after U.S. officials raised security concerns.

4 min readRead the linked source
Source-provided image accompanying Tencent releases Hy4 visual AI model as $7 b Oracle chip‑leasing deal falls apart
Source referenceSource recorded
Publisher
tradingview.com
Source type
Linked source — primary-source status has not been established.
ContextUnderstand this in 60 seconds

Key terms

Robustness
A model's ability to maintain performance under noise, shifts, or adversarial inputs.
Weight
A learned numeric value that scales signals passing through a neural network.
Bias
A consistent pattern of error or unfairness in data or model behavior.

What happened

Tencent Holdings Ltd announced the release of Hy4, an open‑source visual artificial‑intelligence model built to interpret video input and issue commands to a computer. The model is marketed for tasks such as operating Blender, CAD applications, and other productivity software. Tencent said the model weights and a detailed safety‑testing report will be published later, and that pricing for the upcoming Hy4 Preview remains unchanged. In the same weekly recap, the outlet reported that a previously reported $7 b chip‑leasing deal between Tencent and Oracle was cancelled after U.S. officials expressed concerns and Guangdong provincial officials urged Tencent to rely on domestic chips.

Tencent’s Hy4 model is described as a visual AI system that can process video streams and generate mouse‑clicks, keyboard inputs, and other commands to operate software such as Blender, a 3‑D creation suite, and computer‑aided design (CAD) programs. The company framed the release as a step toward "computer control" AI, emphasizing use cases in creative and engineering domains. While the model weights are not yet publicly available, Tencent indicated they will be released soon, and that the pricing for the Hy4 Preview tier remains the same as previously announced.

The announcement also referenced a safety‑testing report that Tencent plans to publish later. According to Reuters, the report will detail how the model was evaluated for defensive cybersecurity tasks and how offensive capabilities were deliberately omitted. This aligns with Tencent’s positioning of Hy4 as a "defensive" AI tool for vulnerability patching and other security‑focused applications.

Separately, the TradingView recap noted that a $7 b chip‑leasing arrangement with Oracle, which had been reported in earlier weeks, was called off. The cancellation followed concerns raised by U.S. officials about the potential for advanced AI chips to be used in ways that could threaten national security, as well as pressure from Guangdong provincial authorities urging Tencent to prioritize domestically produced chips.

Source details: tradingview.com

Why it matters

Hy4 represents one of the first openly available visual‑AI systems that can directly manipulate a computer’s graphical interface, a capability that could lower the barrier for developers to embed AI‑driven automation in design and engineering tools. By positioning the model as open‑source, Tencent invites community scrutiny of its safety mechanisms, which it claims exclude offensive capabilities and focus on defensive cybersecurity tasks such as vulnerability patching. The cancellation of the Oracle chip‑leasing agreement signals heightened geopolitical pressure on Chinese firms to source AI hardware domestically, potentially reshaping Tencent’s hardware roadmap and affecting the broader AI‑chip market. Together, the model launch and the deal’s termination illustrate the twin forces of rapid AI innovation and escalating regulatory scrutiny.

The open‑source nature of Hy4 could accelerate research and product development in visual AI, a field that has traditionally been dominated by proprietary models from large cloud providers. By making the model publicly available, Tencent enables independent verification of its safety claims and invites contributions that could improve and reduce .

Hy4’s focus on computer‑control tasks opens new possibilities for automating complex software workflows, potentially increasing productivity in industries such as animation, architecture, and engineering. However, the same capability also raises concerns about misuse, for example in automating phishing attacks or illicit data extraction, underscoring the importance of the pending safety report.

The termination of the Oracle chip‑leasing deal highlights the growing friction between Chinese AI firms and U.S. technology policy. Access to high‑performance AI chips is a critical bottleneck for training and deploying large models; losing a $7 b supply line may force Tencent to accelerate domestic chip development or seek alternative partners, influencing the competitive dynamics of the global AI‑chip market.

Interactive Mechanism

Interactive Mechanism: How It Actually Works

Explore the underlying technology behind this development interactively.

Model Parameter Size:8B Parameters
VRAM Required5.5 GBGPU memory footprint
Target HardwareMacBook / Single GPUDeployment tier
Privacy100% Air-GappedLocal device capability
Core takeaway: Small, quantized models (3B–8B) now run directly inside smartphones and laptops with complete data privacy, while mammoth 400B+ models remain the domain of datacenter clusters.
Interactive Concept Check+10 Points
AI Models Explained Quiz

Which component of an AI application is the machine-learning model itself?

What to watch next

Key items to monitor include the release date and licensing terms of Hy4’s model weights, the contents of the forthcoming safety‑testing report, and how quickly developers adopt the model in real‑world workflows. Observers should also track Tencent’s next hardware strategy—whether the company accelerates partnerships with Chinese chip makers, invests in in‑house silicon, or seeks alternative overseas suppliers. Finally, regulatory responses in the United States and China regarding AI‑enabled computer control and cross‑border chip leasing will shape the model’s commercial trajectory.

The timing and licensing details of Hy4’s release: whether Tencent adopts an open‑source license that permits commercial use, and how quickly the community can build on the model.

Contents of the safety‑testing report: the methodology used, any identified limitations, and whether third‑party auditors will review the findings.

Tencent’s hardware strategy post‑Oracle deal: announcements of partnerships with Chinese chipmakers, investments in in‑house silicon, or new leasing arrangements with other foreign vendors.

Regulatory developments in the United States and China concerning AI models that can control computers, especially any export controls or licensing requirements that could affect Hy4’s distribution.

Adoption metrics: early adopters in the CAD and 3‑D graphics ecosystems, and any reported performance benchmarks compared to existing visual‑AI solutions.

Related guides & quizzes

Found this useful?