Qwen3-235B-A22B-Thinking-2507 is a powerful open-source MoE model (22B active) built for deep reasoning. It achieves SOTA results on agentic tasks, supports a 256K context, and is available on Hugging Face and via API.
The Qwen team continues to push the upper limits of Qwen3 series with their latest release.
The new model has a very long name—Qwen3-235B-A22B-Thinking-2507—but its capabilities are incredibly strong. This model has SOTA results for open models in core reasoning areas like coding (LiveCodeBench) and math (AIME25), making it competitive with top-tiers like Gemini-2.5 Pro.
The best part is you don't need a complex setup to see it in action. You can experience it directly in Qwen Chat.
@zaczuo Exciting to see Qwen3 launch! Curious – do you see Qwen models becoming strong alternatives for small teams who want more control vs OpenAI/Anthropic?
I run a Medium blog (9K+ monthly views) on AI tools and would love to feature this soon.
Wow, a model that helps you actually *think deeper* OR just get things done faster? That’s honestly genius, ngl. Big props to the Qwen team for this one!
Congrats looking forward to using it, albeit I wasn't too impressed with the last update six months back. Hopefully this one has the firepower to match Moonshot AI's Kimi K2
Replies
Hi everyone!
The Qwen team continues to push the upper limits of Qwen3 series with their latest release.
The new model has a very long name—Qwen3-235B-A22B-Thinking-2507—but its capabilities are incredibly strong. This model has SOTA results for open models in core reasoning areas like coding (LiveCodeBench) and math (AIME25), making it competitive with top-tiers like Gemini-2.5 Pro.
The best part is you don't need a complex setup to see it in action. You can experience it directly in Qwen Chat.
@zaczuo Exciting to see Qwen3 launch! Curious – do you see Qwen models becoming strong alternatives for small teams who want more control vs OpenAI/Anthropic?
I run a Medium blog (9K+ monthly views) on AI tools and would love to feature this soon.
BestPage.ai
Wow, a model that helps you actually *think deeper* OR just get things done faster? That’s honestly genius, ngl. Big props to the Qwen team for this one!
Congrats looking forward to using it, albeit I wasn't too impressed with the last update six months back. Hopefully this one has the firepower to match Moonshot AI's Kimi K2
This is a very good model. Our team was super impressed in its reasoning capabilities within the cyber space.
Benchmark results look pretty intriguing, can't wait to give it a go. Great work.