News
GroqCloud Adds Support for Alibaba’s Qwen3-32B Language Model
- By John K. Waters
- 06/29/2025
Groq has expanded its cloud AI offerings with the integration of Qwen3-32B, a dense 32.8 billion parameter language model developed by Alibaba’s Qwen team.
The new deployment brings multilingual reasoning and conversation capabilities to GroqCloud, a notable step toward making large language models more accessible for production-grade workloads.
Qwen3 32B supports more than 100 languages and dialects, and is optimized for both complex reasoning and dialogue. Alibaba’s latest model introduces a dual-mode system for toggling between thinking and non-thinking states, a feature designed to optimize performance in varied application contexts.
According to independent benchmarks conducted by Artificial Analysis, Groq’s implementation of Qwen3-32B achieves an inference speed of approximately 535 tokens per second.
Groq’s integration of the model includes support for the full 131,072-token context window—a technical capability that distinguishes Groq from other inference providers. This makes it possible for developers to move beyond prototypes and build full-scale, memory-intensive applications on the platform.
What is Groq?
Groq is a U.S.-based AI hardware and cloud inference company known for high-speed, low-latency deployment of large language models. The company offers inference services through its GroqCloud platform, focusing on delivering real-time AI applications that operate at production scale. Its architecture is purpose-built for deterministic throughput and cost-efficiency, even with models requiring extended context windows.
Groq is offering Qwen3-32B at $0.29 per million input tokens and $0.59 per million output tokens. Access is available through GroqChat, the GroqCloud Developer Console, or via API with the model ID qwen/qwen3-32b.
The release follows a trend among cloud providers to support increasingly capable open-source models, particularly those offering multilingual and reasoning capabilities. Qwen3-32B has outperformed prior Qwen iterations, including Qwen2.5 and QwQ, across tasks such as code generation, mathematical problem-solving, and commonsense reasoning.
Groq’s move also underscores a growing demand for large-scale, cost-efficient inference infrastructure capable of supporting advanced language tasks in real-world environments.
About the Author
John K. Waters is the editor in chief of a number of Converge360.com sites, with a focus on high-end development, AI and future tech. He's been writing about cutting-edge technologies and culture of Silicon Valley for more than two decades, and he's written more than a dozen books. He also co-scripted the documentary film Silicon Valley: A 100 Year Renaissance, which aired on PBS. He can be reached at [email protected].