Announcing Qwen 3, now on GMI Cloud

Today we’re excited to announce that Qwen 3 32B and Qwen 3 235B are now available on GMI Cloud’s US-based inference clusters with global deployment support taking advantage of our datacenters around the globe.

Built by Alibaba’s Qwen team and open-sourced under the permissive Apache 2.0 license, Qwen 3 models represent a new leap forward in open LLM performance, flexibility, and multilingual accessibility. And now, for the first time, developers can deploy these models instantly on high-availability, low-latency infrastructure in the USA backed by GMI Cloud’s purpose-built AI stack.

Why Qwen 3 Matters

The flagship Qwen 3 235B-A22B model boasts 235 billion total parameters (22B activated), and rivals the performance of models like Gemini 2.5 Pro and Grok-3 in STEM, coding, long-context tasks, and multilingual reasoning.

Meanwhile, the smaller Qwen 3 32B model offers elite performance at a lighter footprint and lower latency—ideal for production inference at scale.

Key innovations include:

• Hybrid Thinking Modes — Switch between "thinking" (step-by-step reasoning) and "non-thinking" (rapid-response) modes dynamically, depending on task complexity and budget constraints.
  
  
• Massive Context Windows — With up to 128K tokens, Qwen 3 models can handle longer documents, more detailed instructions, and sustained multi-turn conversations.
  
  
• Multilingual Mastery — With support for 119 languages and dialects, Qwen 3 is among the most globally accessible models available today.
  
  
• Agentic-Ready — Optimized for tool use, code execution, and compatibility with emerging agent standards like MCP (Multi-Agent Capability Protocol).
  
  

What This Unlocks for Developers

Qwen 3's hybrid thinking, massive context length, and multilingual fluency create new opportunities for AI developers that simply weren't practical before:

• Dynamic cost-quality tradeoffs: Fine-tune if "thinking" is needed—balancing speed, depth, and cost according to your task.
• International deployment: Build multilingual applications that seamlessly serve users in over 100 languages with native fluency, without needing external translation layers.
• Long-form reasoning: Handle inputs like technical documents, legal contracts, or research papers in a single pass, maintaining nuanced understanding across 128K-token sequences.
• Tool-augmented agents: Build agents that can reason, plan, and interact with APIs and services intelligently, natively supporting tool-calling workflows through MCP integrations.

Real-world use cases now within reach:

• Launch a multilingual support agent that reasons through complex product manuals without needing separate translation pipelines.
• Deploy a global customer service assistant that switches between fast-response mode and deep reasoning depending on user queries.
• Build AI research copilots that analyze full research papers and technical documents in a single session, using full 128K-token context windows.
• Create tool-augmented agents that dynamically interact with APIs, databases, and workflows, powered by native MCP support.
• Develop adaptive agents that toggle between fast interaction and deep thinking modes depending on system load or user preference.

Amplifying what you can do with Qwen

• Customize deployments using our Inference Engine—adjust latency, throughput, and scaling parameters easily to meet specific application needs.
• Optimize resource usage with Cluster Engine—balance GPU allocation dynamically for maximum efficiency and predictable costs.
• Deploy globally with our multi-region infrastructure—giving you the ability to serve users close to their geographic location and fully leverage Qwen 3's multilingual capabilities.
• Scale flexibly by distributing workloads across multiple GPUs—perfect for high-volume, low-latency, or long-context AI applications.

Before Qwen 3, delivering scalable multilingual agents, reasoning engines, or cost-optimized AI applications meant stitching together multiple models or relying on proprietary platforms. Now, it’s open-source—and production-ready !—on GMI Cloud.

Why GMI Cloud

GMI Cloud is purpose-built for the AI workloads of today and tomorrow:

• Inference-Optimized Clusters — Tuned for high-throughput, low-latency large model serving.
  
  
• Transparent Pricing — Simple, predictable billing without hidden fees.
  
  
• Instant API Access — Launch OpenAI-compatible APIs through frameworks like vLLM and SGLang with minimal setup.
  
  
• Enterprise-Grade Reliability — High availability, secure deployments, and scalable capacity as your needs grow.
  
  

Whether you're running autonomous agents, building a multilingual co-pilot, or researching new AI behaviors, Qwen 3 is now just a few clicks away.

Get Started

Ready to build agents, copilots, or next-gen AI products?

Spin up Qwen 3 32B and 235B today on GMI Cloud’s Inference Engine—with flexible scaling, API simplicity, and no surprises.

Read Qwen's blog announcement.

Build faster, think deeper—with Qwen 3 on GMI Cloud.