Qwen3 235B A22B Thinking 2507 needs about 144.7 GB of memory at Q4_K_M (a 142.6 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 24 quants, from 0.5 GB to 249.9 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 6.3 GB.
Qwen3 235B A22B Thinking 2507 – a large‑scale reasoning model
Qwen3 235B A22B Thinking 2507 is a 235B‑parameter causal language model with 22B active experts. It supports a native context length of 262,144 tokens and is optimized for complex reasoning tasks. The model is released under the apache‑2.0 license.
Good at
- Logical reasoning
- Mathematics
- Scientific problem solving
Limits
- No vision support
- Requires large VRAM for full precision
- Only thinking mode is supported
Best for
- Academic research requiring deep reasoning
- Complex coding assistance
- Long‑form analytical writing
Who it’s for. Researchers and developers who need a high‑capacity reasoning model. The model is suitable for environments with sufficient GPU memory and a need for extended context handling.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.5 GB | 2.6 GB | Fits | Get file → | |
| IQ1_M | 53.2 GB (2 parts) | 55.3 GB | Too big | Get file → | |
| IQ2_XS | 65.6 GB (2 parts) | 67.7 GB | Too big | Get file → | |
| IQ2_S | 66.1 GB (2 parts) | 68.2 GB | Too big | Get file → | |
| IQ2_M | 78.6 GB (2 parts) | 80.6 GB | Too big | Get file → | |
| Q2_K | 82.7 GB (3 parts) | 84.8 GB | Too big | Get file → | |
| Q2_K_L | 83.3 GB (3 parts) | 85.4 GB | Too big | Get file → | |
| IQ3_XXS | 93.1 GB (3 parts) | 95.1 GB | Too big | Get file → | |
| IQ3_XS | 96.9 GB (3 parts) | 99.0 GB | Too big | Get file → | |
| Q3_K_S | 102.4 GB (3 parts) | 104.5 GB | Too big | Get file → | |
| IQ3_M | 107.3 GB (3 parts) | 109.4 GB | Too big | Get file → | |
| Q3_K_M | 107.3 GB (3 parts) | 109.4 GB | Too big | Get file → | |
| Q3_K_L | 111.2 GB (3 parts) | 113.2 GB | Too big | Get file → | |
| Q3_K_XL | 111.7 GB (3 parts) | 113.8 GB | Too big | Get file → | |
| IQ4_XS | 125.9 GB (4 parts) | 128.0 GB | Too big | Get file → | |
| IQ4_NL | 133.1 GB (4 parts) | 135.2 GB | Too big | Get file → | |
| Q4_0 | 135.0 GB (4 parts) | 137.1 GB | Too big | Get file → | |
| Q4_K_S | 137.7 GB (4 parts) | 139.8 GB | Too big | Get file → | |
| Q4_K_M | 142.6 GB (4 parts) | 144.7 GB | Too big | Get file → | |
| Q4_1 | 147.3 GB (4 parts) | 149.4 GB | Too big | Get file → | |
| Q5_K_S | 162.0 GB (5 parts) | 164.0 GB | Too big | Get file → | |
| Q5_K_M | 166.9 GB (5 parts) | 169.0 GB | Too big | Get file → | |
| Q6_K | 193.1 GB (5 parts) | 195.1 GB | Too big | Get file → | |
| Q8_0 | 249.9 GB (7 parts) | 252.0 GB | Too big | Get file → |
For 8 GB VRAM use Q4_0 (file size 134.99 GB, ~137.1 GB memory at 8 k context). For 16 GB VRAM use Q4_K_M (142.65 GB, ~144.7 GB memory). For 32 GB VRAM use Q5_K_M (166.89 GB, ~169 GB memory). Estimates are approximate.
What GPU do you need for Qwen3 235B A22B Thinking 2507?
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: the best quant that fits is the main file (0.5 GB file, about 2.6 GB in use).
- 6 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 2.6 GB in use).
- 8 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 2.6 GB in use).
- 12 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 2.6 GB in use).
- 16 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 2.6 GB in use).
- 24 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 2.6 GB in use).
- 32 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 2.6 GB in use).
- 48 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 2.6 GB in use).
Which Qwen3 235B A22B Thinking 2507 quant should you choose?
Q4_K_M is the usual starting point for Qwen3 235B A22B Thinking 2507: it balances file size and quality well. Here it is a 142.6 GB file.
The Q8_0 file is 249.9 GB, about 1.8 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is the first one in the table at 0.5 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Qwen3 235B A22B Thinking 2507
How much VRAM does Qwen3 235B A22B Thinking 2507 need?
About 144.7 GB at Q4_K_M with an 8K context window. That is the 142.6 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 235B A22B Thinking 2507 on an 8 GB GPU?
Yes, with the smaller files: the 0.5 GB file needs about 2.6 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Qwen3 235B A22B Thinking 2507 GGUF?
It depends on the quant: file is 0.5 GB, IQ1_M is 53.2 GB, IQ2_XS is 65.6 GB, IQ2_S is 66.1 GB, IQ2_M is 78.6 GB, Q2_K is 82.7 GB and 18 more.
What context length does Qwen3 235B A22B Thinking 2507 support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 1.6 GB at 8K and 6.3 GB at 32K.
Where can I download Qwen3 235B A22B Thinking 2507 GGUF?
The files are on Hugging Face in the bartowski/Qwen_Qwen3-235B-A22B-Thinking-2507-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 235B A22B Thinking 2507 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 235B A22B Thinking 2507 on 8 GB VRAM?
Use the Q4_0 quantization; it requires about 137.1 GB memory at 8 k context, which is beyond 8 GB VRAM, so it is not feasible.
Does the model support image input?
No, the model is text‑only and does not support vision capabilities.
What license applies to this model?
The model is released under the apache‑2.0 license.