Qwen3-4B-GGUF
Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Built upon extensive training, Qwen3 delivers groundbreaking advancements in reasoning, instruction-following, agent capabilities, and multilingual support, with the following key features:
Params
—
Context
—
Downloads 30d
280 K
Likes
135
Download history
daily snapshots · 10 days288 K277 K
Jul 28Jul 31Aug 3Aug 6
Can you run it?
Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.
| File | Quant | Size | Est. VRAM | Verdict on RTX 4090 · 24 GB |
|---|---|---|---|---|
| Qwen3-4B-Q4_K_M.gguf | Q4_K_M | 2.5 GB | 3.2 GB | ✅ Runs comfortably |
| Qwen3-4B-Q5_0.gguf | Q5_0 | 2.8 GB | 3.6 GB | ✅ Runs comfortably |
| Qwen3-4B-Q5_K_M.gguf | Q5_K_M | 2.9 GB | 3.7 GB | ✅ Runs comfortably |
| Qwen3-4B-Q6_K.gguf | Q6_K | 3.3 GB | 4.1 GB | ✅ Runs comfortably |
| Qwen3-4B-Q8_0.gguf | Q8_0 | 4.3 GB | 5.2 GB | ✅ Runs comfortably |
Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.
Run it
copy-paste, exact tags checked against the Hub$ ollama run qwen3-4b-gguf # pin the quantization explicitly $ ollama run qwen3-4b-gguf-q4_k_m
$ huggingface-cli download Qwen/Qwen3-4B-GGUF-GGUF \
Qwen3-4B-Q4_K_M.gguf --local-dir .
$ llama-cli -m Qwen3-4B-Q4_K_M.gguf \
-c 8192 -ngl 99 -t 8 --color
$ curl -s https://aimodelscomparison.com/api/v1/models/qwen3-4b-gguf
{
"hf_id": "Qwen/Qwen3-4B-GGUF",
"params_b": null,
"context_length": null,
"license": { "id": "apache-2.0", "commercial": "yes" },
"downloads_30d": 280207,
"vram_estimates": [
{ "quant": "Q4_K_M", "gb": 3.2 },
{ "quant": "Q5_0", "gb": 3.6 }
],
"updated_at": "2026-07-28T18:06:52Z"
}
Specifications
- Licence
- apache-2.0
- First seen on the Hub
- 2025-05-05
- Base model
- Qwen3-4B
- Training datasets
- undisclosed
- Added to our catalog
- 2026-07-28
Sponsored · GPU cloud
Not enough VRAM?
Spin up a 24 GB L4 instance in 40 seconds. $0.44/hr.