You asked, we listened, and Qwen3-32B with its support of 100+ languages and 128K context window is NOW LIVE on Groq for real-time speed! 
What’s so exciting about Qwen 3? Let’s see...
Hybrid Thinking Modes 🧠
You can control how much the model "thinks" before responding, which means you can control your computational budget based on what you need. 🤯 We’ve introduced the brand new reasoning_effort parameter to Groq API that you can set to none to disable reasoning for quick responses, or default to enable step-by-step reasoning for complex problems.
This is HUGE for those of us who complained about how chatty Qwen’s QwQ-32B model was. Need short outputs and less tokens consumed for simple tasks? No problem. Just enable or disable the model’s reasoning effort whenever you want.
Smol but Mighty 
Qwen3-32B performs as well as their previous-generation 72B model while being significantly smaller. The efficiency gains are incredible - major savings in inference and training costs while delivering better performance that we can pass down to you with our pricing of $0.29/1M input tokens and $0.59/1M output tokens.
True Multilingual Support 
We're talking 100+ languages with actual comprehension, not just basic translation. Arabic, Bengali, Farsi, Hindi, Tamil, Turkish, Japanese, Korean, and so many more.
128K Context Window 
All running at Groq speed. Seriously, only we can deliver this combination of massive context with real-time inference.
Ready to experience it for yourself?
You can try it right now on both our Free Tier and Developer Tier via our console playground or via Groq API: https://console.groq.com/playground?model=qwen/qwen3-32b
Check out our reasoning docs for more info on reasoning models including Qwen3-32B: https://console.groq.com/docs/reasoning
Also, shoutout to our summer intern, Krish, who built Project Linguist to share with you all - an open-source demo that generates content across all 100+ languages using Qwen3-32B on Groq! 🤩 Pretty awesome example of what's possible. Clone the repo to try and customize for your own needs: https://github.com/KrishDesai/project-linguist
I'm genuinely curious to read your feedback on Qwen3-32B with its hybrid reasoning mode and I'd love to hear which models or features you'd be most excited about seeing on Groq. You know your feedback shapes what we prioritize, so keep those feature requests coming!
And, as always, happy building!