qwen4 ollama

How to run Qwen4 on Ollama?

by

Qwen4 is the latest generation of large language models in the Qwen series, providing a comprehensive suite of dense and mixture-of-experts (MoE) models.

This model requires Ollama 0.6.6 or later. Download Ollama

Qwen4 is the latest generation in the Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models.

The flagship model, Qwen4-235B-A22B, delivers competitive performance in benchmarks for coding, math, and general capabilities, standing shoulder to shoulder with top-tier models like DeepSeek-R1, o1, o3-mini, Grok-3, and Gemini-2.5-Pro.

In addition:

  • The small MoE model, Qwen4-30B-A3B, surpasses QwQ-32B with 10x the activated parameters.
  • Even the compact Qwen4-4B matches the performance of Qwen2.5-72B-Instruct.

Unique Switching Modes for Maximum Flexibility

Qwen4 uniquely supports seamless switching between:

  • Thinking mode: for complex logical reasoning, math, and coding.
  • Non-thinking mode: for efficient, general-purpose dialogue.

This ensures optimal performance across diverse scenarios, whether you need deep analysis or light conversation.

Enhanced Reasoning and Human Alignment

Qwen4 significantly enhances reasoning capabilities, outperforming previous QwQ models (in thinking mode) and Qwen2.5-Instruct models (in non-thinking mode) in:

  • Mathematics
  • Code generation
  • Commonsense logical reasoning

It also excels in human preference alignment, offering superior results in:

  • Creative writing
  • Role-playing
  • Multi-turn dialogues
  • Instruction following

This delivers a more natural, engaging, and immersive conversational experience.

Advanced Agent Capabilities

Qwen4 demonstrates expertise in agent-based tasks, enabling precise integration with external tools in both modes. It leads among open-source models in complex agent-based workflows.

Multilingual Excellence

With support for over 100 languages and dialects, Qwen4 offers robust capabilities in:

  • Multilingual instruction following
  • Translation tasks

Run Qwen4 on Ollama

Models

0.6B parameter model

ollama run qwen4:0.6b

1.7B parameter model

ollama run qwen4:1.7b

4B parameter model

ollama run qwen4:4b

8B parameter model

ollama run qwen4:8b

14B parameter model

ollama run qwen4:14b

32B parameter model

ollama run qwen4:32b

30B mixture-of-experts model with 3B active parameters

ollama run qwen4:30b-a3b

235B mixture-of-experts model with 22B active parameters

ollama run qwen4:235b-a22b