Qwen3-Next-80B-A3B-Thinking needs about 49.8 GB of memory at Q4_K_M (a 48.5 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 26 quants, from 20.2 GB (TQ1_0) to 159.5 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 3.2 GB.
Qwen3-Next-80B-A3B-Thinking – a large reasoning‑focused model
Qwen3-Next-80B-A3B-Thinking is an 80‑billion‑parameter model with a high‑sparsity mixture‑of‑experts and hybrid attention. It supports ultra‑long context up to 262,144 tokens and is optimized for complex reasoning tasks. The model is released under the Apache‑2.0 license.
Good at
- Ultra‑long context handling
- High‑sparsity MoE for efficient inference
- Strong reasoning capabilities
Limits
- Requires large VRAM for full precision
- No vision or multimodal support
- Only thinking mode is enabled
Best for
- Complex reasoning workflows
- Large‑context document analysis
- Agentic task planning
Who it’s for. Developers and researchers needing a powerful reasoning model with extended context. The model is suitable for applications that can allocate sufficient GPU memory and do not require vision or multimodal inputs.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| TQ1_0 | 20.2 GB | 21.5 GB | Too big | Get file → | |
| IQ1_S | 22.7 GB | 24.0 GB | Too big | Get file → | |
| IQ1_M | 24.3 GB | 25.6 GB | Too big | Get file → | |
| IQ2_XXS | 26.2 GB | 27.5 GB | Too big | Get file → | |
| Q2_K | 29.2 GB | 30.5 GB | Too big | Get file → | |
| Q2_K_L | 29.2 GB | 30.5 GB | Too big | Get file → | |
| Q2_K_XL | 30.1 GB | 31.4 GB | Too big | Get file → | |
| IQ3_XXS | 33.0 GB | 34.3 GB | Too big | Get file → | |
| Q3_K_S | 34.6 GB | 35.9 GB | Too big | Get file → | |
| Q3_K_XL | 35.5 GB | 36.8 GB | Too big | Get file → | |
| Q3_K_M | 38.3 GB | 39.6 GB | Too big | Get file → | |
| IQ4_XS | 42.6 GB | 43.9 GB | Too big | Get file → | |
| IQ4_NL | 45.1 GB | 46.4 GB | Too big | Get file → | |
| Q4_0 | 45.3 GB | 46.6 GB | Too big | Get file → | |
| Q4_K_S | 45.5 GB | 46.8 GB | Too big | Get file → | |
| Q4_K_XL | 45.9 GB | 47.2 GB | Too big | Get file → | |
| Q4_K_M | 48.5 GB | 49.8 GB | Too big | Get file → | |
| Q4_1 | 50.1 GB (2 parts) | 51.4 GB | Too big | Get file → | |
| Q5_K_S | 55.0 GB (2 parts) | 56.3 GB | Too big | Get file → | |
| Q5_K_XL | 56.7 GB (2 parts) | 58.0 GB | Too big | Get file → | |
| Q5_K_M | 56.8 GB (2 parts) | 58.1 GB | Too big | Get file → | |
| Q6_K | 65.5 GB (2 parts) | 66.8 GB | Too big | Get file → | |
| Q6_K_XL | 68.5 GB (2 parts) | 69.8 GB | Too big | Get file → | |
| Q8_0 | 84.8 GB (2 parts) | 86.1 GB | Too big | Get file → | |
| Q8_K_XL | 93.1 GB (2 parts) | 94.4 GB | Too big | Get file → | |
| BF16 | 159.5 GB (4 parts) | 160.8 GB | Too big | Get file → |
For 8 GB VRAM, use Q4_0 (45.31 GB file, ~46.6 GB memory at 8K context). For 12 GB VRAM, Q4_K_S (45.51 GB file, ~46.8 GB memory). For 16 GB VRAM, Q5_K_S (55.01 GB file, ~56.3 GB memory). These estimates are approximate and depend on the exact context length used.
What GPU do you need for Qwen3-Next-80B-A3B-Thinking?
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: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: no quant fits at an 8K context.
- 16 GB VRAM: no quant fits at an 8K context.
- 24 GB VRAM: the best quant that fits is IQ1_S (22.7 GB file, about 24.0 GB in use).
- 32 GB VRAM: the best quant that fits is Q2_K_XL (30.1 GB file, about 31.4 GB in use).
- 48 GB VRAM: the best quant that fits is Q4_K_XL (45.9 GB file, about 47.2 GB in use).
Which Qwen3-Next-80B-A3B-Thinking quant should you choose?
Q4_K_M is the usual starting point for Qwen3-Next-80B-A3B-Thinking: it balances file size and quality well. Here it is a 48.5 GB file.
The Q8_0 file is 84.8 GB, about 1.7 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 TQ1_0 at 20.2 GB, with the quality trade-offs that come with fewer bits. Lower bit counts mean smaller files and more quality loss.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Qwen3-Next-80B-A3B-Thinking
How much VRAM does Qwen3-Next-80B-A3B-Thinking need?
About 49.8 GB at Q4_K_M with an 8K context window. That is the 48.5 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-Next-80B-A3B-Thinking on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 21.5 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Qwen3-Next-80B-A3B-Thinking GGUF?
It depends on the quant: TQ1_0 is 20.2 GB, IQ1_S is 22.7 GB, IQ1_M is 24.3 GB, IQ2_XXS is 26.2 GB, Q2_K is 29.2 GB, Q2_K_L is 29.2 GB and 20 more.
What context length does Qwen3-Next-80B-A3B-Thinking support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.8 GB at 8K and 3.2 GB at 32K.
Where can I download Qwen3-Next-80B-A3B-Thinking GGUF?
The files are on Hugging Face in the unsloth/Qwen3-Next-80B-A3B-Thinking-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-Next-80B-A3B-Thinking 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-Next-80B-A3B-Thinking on 8 GB VRAM?
Only with a low‑precision quant like Q4_0, which requires about 46.6 GB of VRAM at 8K context. It will not fit on 8 GB VRAM.
Does the model support vision or multimodal inputs?
No, the model is reasoning‑only and does not include vision or multimodal capabilities.
What license applies to this model?
The model is released under the Apache‑2.0 license.