Skip to main content
PromptQuorumBuilt for humans. Structured for AI.

Can You Run Qwen 3 on Ollama?

Can You Run Qwen 3 on Ollama?
Read in:
🇺🇸en🇩🇪de🇫🇷fr🇯🇵ja🇨🇳zh🇪🇸es🇧🇷pt🇸🇦ar🇰🇷ko

Quick Answer

Yes — Ollama supports all Qwen 3 model sizes from 0.6B to 32B, plus the 30B-A3B MoE variant, with native tool calling via the standard API, needing only a single command like ollama run qwen3:8b. The 8B model needs ~6 GB VRAM at Q4.

  • ▸ollama run qwen3:0.6b — fits in 1 GB VRAM
  • ▸ollama run qwen3:8b — needs ~6 GB VRAM
  • ▸ollama run qwen3:32b — needs ~21 GB VRAM
OllamaIntermediate

Key Takeaways

  • ✓Ollama supports all Qwen 3 sizes: 0.6B, 1.7B, 4B, 8B, 14B, and 32B, plus the 30B-A3B MoE variant
  • ✓Pull any size with <code>ollama run qwen3:8b</code> — replace the tag with your target size
  • ✓The 8B model needs ~6 GB VRAM at Q4 and runs at ~20 tok/s on a mid-range GPU
  • ✓Qwen 3 supports tool calling natively via the standard Ollama API — no custom Modelfile required

Yes — Here's What's Available

Ollama supports all major Qwen 3 model sizes from 0.6B to 32B, plus the 30B-A3B MoE variant. Pull any size with a single command: ollama run qwen3:8b. Replace 8b with 0.6b, 1.7b, 4b, 14b, 32b, or 30b-a3b for other sizes.

Each size is available in multiple quantizations. Q4_K_M is the default and recommended starting point — it delivers the best quality-to-file-size ratio. Q8_0 is available for 8B and 14B if you have the VRAM headroom.

Tool calling is supported natively on all Qwen 3 sizes via the standard Ollama API. No custom Modelfile or special prompt template is required.

ollama run qwen3:8b

Which Qwen 3 Size to Pick

The right Qwen 3 size depends entirely on available VRAM. For most users on a mid-range GPU (6–8 GB VRAM), the 8B model at Q4_K_M is the practical choice — it needs ~6 GB and runs at ~20 tok/s.

The 14B model at Q4 is the recommended coding tier: it outperforms the 8B on code generation and fits comfortably in 10–12 GB VRAM. For a full comparison of Qwen 3 coding performance versus other local models, see the guide to running Qwen locally in 2026.

VRAMQwen 3 SizeBest For
< 4 GB0.6B / 1.7BEdge devices, testing, CPU-only
4–6 GB4BBudget GPU or low-RAM CPU
6–12 GB8B / 14BGeneral use and coding
12–24 GB32B / 30B-A3B (MoE)High-quality coding and reasoning

Quick Answers About Qwen 3 on Ollama

How do I install Qwen 3 on Ollama?▾
Run ollama run qwen3:8b in a terminal. Ollama downloads the model automatically on first run. Replace 8b with your target size: 0.6b, 1.7b, 4b, 14b, 32b, or 30b-a3b.
Is Qwen 3 better than Llama 3 for coding?▾
For coding: yes, Qwen3 14B outperforms Llama 3 8B on HumanEval benchmarks. For general conversation at the 8B tier: Llama 3 8B remains competitive. For the current top Ollama picks across all tasks, see the best Ollama models right now.
Does Qwen 3 support tool calling on Ollama?▾
Yes. Qwen 3 supports function and tool calling natively via the standard Ollama API. No custom Modelfile or special configuration is required — it works with any client that supports the Ollama tool-use format.
What is the largest Qwen 3 model you can run locally?▾
The largest dense Qwen 3 model is 32B, which needs ~21 GB of VRAM at Q4 — it fits on a single RTX 3090 or RTX 4090. For faster inference at similar quality, the 30B-A3B MoE variant needs only ~19 GB VRAM and runs noticeably faster because just a fraction of its parameters activate per token. Apple M-series Macs with 32+ GB unified memory run either comfortably.

Want the full breakdown?

Read the complete guide →

Related Prompt Bites

← Back to Prompt Bites