OpenAI has officially released its two newest open-weight models: GPT-OSS-120B and GPT-OSS-20B. This marks a significant milestone in promoting open, transparent, and accessible AI for the global developer community.

Reinforcing our leadership in the AI trend, we have promptly integrated these two state-of-the-art models into VNG Cloud Model as a Service, a key component of VNG Cloud AI Stack.

VNG Cloud Model as a Service brings together leading AI models from top global providers such as OpenAI, Google, Anthropic, and DeepSeek, along with specialized models tailored for Vietnamese language and industry-specific use cases. This enables enterprises to deploy AI solutions quickly and flexibly, customized to their unique needs.

High performance on optimized infrastructure

Unlike dense models such as GPT-3.5 or Mistral 7B, the GPT-OSS models are built on Mixture of Experts (MoE) architecture, which activates only a subset of parameters at each computation step—significantly reducing hardware requirements while maintaining high performance.

• GPT-OSS-120B features 120 billion total parameters, but only activates approximately 5.1 billion per inference step
• GPT-OSS-20B activates 3.6 billion out of its 20 billion parameters

You can easily deploy these models directly via VNG Cloud Model as a Service for fast integration into your existing systems. For more customization and control, models can also be deployed using inference platforms or Kubernetes Service (VKS) with NVIDIA H100 GPUs on VNG Cloud AI Infrastructure.

Beyond reducing resource consumption, GPT-OSS models are also optimized for high throughput, making them ideal for real-time AI services such as chatbots, fast-response APIs, and microservices.

Advanced features that drive real-world AI performance

These latest models from OpenAI are not only strong in natural language understanding and generation but are also packed with features that solve real business challenges.

Use cases that turn AI potential into business

These new models enable enterprises to build smart, domain-specific AI applications quickly and effectively.

For internal chatbots, businesses can leverage structured outputs and tool-calling capabilities to build bots that do more than just answer questions—they can retrieve information, generate reports, and even trigger APIs.

For document automation, GPT-OSS models support advanced data extraction and formatting using Chain-of-Thought (CoT) techniques, streamlining tasks like contract processing, email classification, and report summarization.

With powerful reasoning capabilities and context-aware customization, the models can also serve as intelligent assistants across industries—from healthcare and finance to legal and compliance—offering accurate, industry-specific support.

Go live with AI in just minutes

Built on high-performance NVIDIA H100 GPUs, VNG Cloud Model as a Service lets you access and deploy the latest GPT-OSS models instantly—without the hassle of complex setup or costly infrastructure.

To get started:

• Create an account on the VNG Cloud AI Platform

• Instantly access the models via the portal: gpt-oss-120b or gpt-oss-20b

Start your AI journey today

Unlock the power of AI with GPT-OSS-120B and GPT-OSS-20B, now available on VNG Cloud Model as a Service.

Explore VNG Cloud AI Stack—a comprehensive AI ecosystem that helps businesses deploy smarter, faster, and more cost-effective AI solutions. Whether you're an AI expert or just getting started, we have the right solution for you.