Green Fern
Green Fern

Adaptive Inference

AI Infrastructure

Large Language Models

Foundation models

AI models

LLMs

Generative AI

Open Source LLMs

Qwen3-235B-A22B is now live on kluster.ai

Qwen3-235B-A22B is now live on kluster.ai

By Anjin Stewart-Funai

May 1, 2025

Powerful agent support. State-of-the-art coder with optional reasoning. Smarter workflows.

The Qwen family just got a major upgrade, and we’re excited to bring it to you on kluster.ai.

Qwen3-235B-A22B is now available for inference, and it’s built to handle the most complex agent workflows, competitive coding tasks, and optional multi-step reasoning - all while offering blazing-fast performance thanks to its MoE architecture.

Let’s break down what makes this model stand out.

🧠 Built for agentic workflows - From the ground up

Qwen3 isn’t just tool-aware - it’s tool-native.

It comes with first-class support for function calling, API integration, and multi-step orchestration, making it ideal for:

• Retrieval-Augmented Generation (RAG)

• Code execution pipelines

• Multi-tool agent chains

It’s also fully compatible with the Model Context Protocol (MCP), allowing developers to chain together complex reasoning steps and tool calls with ease.

💻 A coder that rivals closed models

Qwen3-235B-A22B leads in open-weight performance for tool use and coding, so whether you’re debugging, generating, or optimizing code, it has you covered.

#1 open model on the BFCL benchmark (tool use), outperforming GPT-4.5

High scores on LiveCodeBench & CodeForces, excelling at real-world coding challenges

• Performs on par with top proprietary models like OpenAI o1, Gemini 2.5 Pro, Grok 3 Beta, and o3-mini

🧪Dynamic reasoning with /think and /nothink

Qwen3 is the first model to support prompt-level reasoning control - making it ideal for agents that need to adapt to different task types in real time:

• /think – triggers step-by-step reasoning, great for complex or high-stakes tasks

• /nothink – skips reasoning for fast, to-the-point outputs

🌍 Multilingual Mastery

Qwen3 was trained on 119 languages, giving it remarkable fluency and context handling across:

• Full codebases

• Technical documents

• Multilingual prompts and datasets

If your agents operate in global contexts or multilingual environments, Qwen3 is a powerful ally.

🔐 Your Data Stays Yours

As always, when you use our OpenAI compatible real time API on kluster.ai:

• Your prompts and outputs are never stored

• You maintain full ownership over your work

Try Qwen3 Now

Qwen3-235B-A22B is live on kluster.ai, available for both API and UI access. Whether you’re building autonomous agents, evaluating open models, or looking for a reliable coding assistant - this one’s worth trying.