Qwen3 4B Thinking 2507 Unsloth needs about 4.2 GB of memory at Q4_K_M (a 2.5 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 9 quants, from 2.4 GB (IQ4_NL) to 4.2 GB (MXFP4_MOE). Longer context needs more: at 32K tokens the KV cache alone adds about 4.8 GB.
Qwen3 4B Thinking 2507 – a 4.02B‑parameter reasoning model
Qwen3 4B Thinking 2507 is a causal language model with 4.02B parameters and 262,144 native context length. It is optimized for complex reasoning, tool usage, and instruction following. The model is released under the apache-2.0 license.
Good at
- Logical and mathematical reasoning
- Tool‑enabled task execution
- Long‑context understanding
Limits
- No vision support
- Requires enable_thinking mode only
- Large VRAM needed for full context
Best for
- Complex problem solving
- Automated reasoning workflows
- Educational tutoring
Who it’s for. Developers and researchers needing a lightweight yet powerful reasoning LLM. The model is suitable for applications that demand deep logical analysis and tool integration.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ4_NL | 2.4 GB | 4.1 GB | Fits | Get file → | |
| MXFP4_MOE | 2.5 GB | 4.2 GB | Fits | Get file → | |
| Q4_K_M | 2.5 GB | 4.2 GB | Fits | Get file → | |
| MXFP4_MOE | 2.5 GB | 4.3 GB | Fits | Get file → | |
| Q5_K | 2.9 GB | 4.6 GB | Fits | Get file → | |
| MXFP4_MOE | 3.1 GB | 4.8 GB | Fits | Get file → | |
| MXFP4_MOE | 3.3 GB | 5.0 GB | Fits | Get file → | |
| IQ4_NL | 3.3 GB | 5.0 GB | Fits | Get file → | |
| MXFP4_MOE | 4.2 GB | 5.9 GB | Fits | Get file → |
For 8 GB VRAM, use IQ4_NL (2.39 GB, 4.1 GB memory at 8k context) or Q4_K_M (2.5 GB, 4.2 GB). If 12 GB VRAM is available, MXFP4_MOE (2.55 GB, 4.3 GB) or Q5_K (2.89 GB, 4.6 GB) provide better quality. Larger models like MXFP4_MOE (3.26 GB, 5 GB) fit 16 GB VRAM setups.
What GPU do you need for Qwen3 4B Thinking 2507 Unsloth?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: no quant fits at an 8K context.
- 6 GB VRAM: the best quant that fits is MXFP4_MOE (4.2 GB file, about 5.9 GB in use).
Which Qwen3 4B Thinking 2507 Unsloth quant should you choose?
Q4_K_M is the usual starting point for Qwen3 4B Thinking 2507 Unsloth: it balances file size and quality well. Here it is a 2.5 GB file.
If memory is tight, the smallest file is IQ4_NL at 2.4 GB, with the quality trade-offs that come with fewer bits. The usual sweet spot between file size and quality, and the quant most people start with.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Qwen3 4B Thinking 2507 Unsloth
How much VRAM does Qwen3 4B Thinking 2507 Unsloth need?
About 4.2 GB at Q4_K_M with an 8K context window. That is the 2.5 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run Qwen3 4B Thinking 2507 Unsloth on an 8 GB GPU?
Yes, with MXFP4_MOE: the 4.2 GB file needs about 5.9 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Qwen3 4B Thinking 2507 Unsloth GGUF?
It depends on the quant: IQ4_NL is 2.4 GB, MXFP4_MOE is 2.5 GB, Q4_K_M is 2.5 GB, MXFP4_MOE is 2.5 GB, Q5_K is 2.9 GB, MXFP4_MOE is 3.1 GB and 3 more.
What context length does Qwen3 4B Thinking 2507 Unsloth support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 1.2 GB at 8K and 4.8 GB at 32K.
Where can I download Qwen3 4B Thinking 2507 Unsloth GGUF?
The files are on Hugging Face in the magiccodingman/Qwen3-4B-Thinking-2507-Unsloth-MagicQuant-Hybrid-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Qwen3 4B Thinking 2507 Unsloth use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run Qwen3 4B Thinking 2507 on 8 GB VRAM?
Yes, the IQ4_NL quant uses 2.39 GB file size and requires about 4.1 GB memory at 8k context, which fits 8 GB VRAM.
Does this model support vision tasks?
No, the model only supports text and reasoning; vision is not included.
What license applies to this model?
The model is released under the apache-2.0 license.