qwen
72B parameters
Commercial OK

Qwen 3 72B

Qwen 3 dense flagship. Reasoning-mode toggle via <think> blocks; same convention as the rest of the Qwen 3 line. Beats Qwen 2.5 72B on reasoning benchmarks at the same parameter count.

License: Apache 2.0·Released Sep 15, 2025·Context: 131,072 tokens

Overview

Qwen 3 dense flagship. Reasoning-mode toggle via <think> blocks; same convention as the rest of the Qwen 3 line. Beats Qwen 2.5 72B on reasoning benchmarks at the same parameter count.

Family & lineage

How this model relates to others in its lineage. Family members share architecture and training-data roots; parent / children edges record direct distillation or fine-tune relationships.

Parent / base model
Qwen 3 32B32B
Workstation

Strengths

  • Reasoning-mode toggle
  • Apache 2.0
  • Strong long-context (128K)

Weaknesses

  • Needs 48GB VRAM at AWQ-INT4 — workstation tier

Quantization variants

Each quantization trades model quality for file size and VRAM. Q4_K_M is the most popular starting point.

QuantizationFile sizeVRAM required
AWQ-INT442.0 GB48 GB

Get the model

HuggingFace

Original weights

huggingface.co/Qwen/Qwen3-72B

Source repository — direct quantization required.

Hardware that runs this

Cards with enough VRAM for at least one quantization of Qwen 3 72B.

Compare alternatives

Models worth comparing

Same parameter band, plus what's one tier above and below — so you can decide what actually fits your hardware.

Frequently asked

What's the minimum VRAM to run Qwen 3 72B?

48GB of VRAM is enough to run Qwen 3 72B at the AWQ-INT4 quantization (file size 42.0 GB). Higher-quality quantizations need more.

Can I use Qwen 3 72B commercially?

Yes — Qwen 3 72B ships under the Apache 2.0, which permits commercial use. Always read the license text before deployment.

What's the context length of Qwen 3 72B?

Qwen 3 72B supports a context window of 131,072 tokens (about 131K).

Source: huggingface.co/Qwen/Qwen3-72B

Reviewed by RunLocalAI Editorial. See our editorial policy for how we research and verify model claims.