INTELLECT 3 needs about 73.4 GB of memory at Q4_K_M (a 71.4 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 27 quants, from 0.2 GB to 113.6 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 6.2 GB.
INTELLECT‑3: 106B Mixture‑of‑Experts reasoning model
INTELLECT‑3 is a 106B parameter Mixture‑of‑Experts model built on glm4moe. It is trained with supervised fine‑tuning and large‑scale reinforcement learning. The model is released under the MIT license.
Good at
- Reasoning
- Large‑scale RL training
- Mixture‑of‑Experts architecture
Limits
- No vision support
- Large memory requirement for full precision
- Requires 131072 context tokens
Best for
- Complex reasoning tasks
- Math and coding challenges
- Research on MoE models
Who it’s for. Developers and researchers needing a high‑capacity reasoning model. The model is open‑source and can be run with llama.cpp or vLLM on compatible hardware.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.2 GB | 2.3 GB | Fits | Get file → | |
| IQ1_S | 35.3 GB | 37.3 GB | Too big | Get file → | |
| IQ1_M | 35.9 GB | 38.0 GB | Too big | Get file → | |
| IQ2_XXS | 37.7 GB | 39.7 GB | Too big | Get file → | |
| IQ2_XS | 40.3 GB | 42.3 GB | Too big | Get file → | |
| IQ2_S | 40.6 GB | 42.7 GB | Too big | Get file → | |
| IQ2_M | 43.2 GB | 45.3 GB | Too big | Get file → | |
| Q2_K | 44.7 GB | 46.8 GB | Too big | Get file → | |
| Q2_K_L | 45.3 GB | 47.4 GB | Too big | Get file → | |
| IQ3_XXS | 48.3 GB | 50.4 GB | Too big | Get file → | |
| IQ3_XS | 49.2 GB | 51.3 GB | Too big | Get file → | |
| Q3_K_S | 51.7 GB (2 parts) | 53.8 GB | Too big | Get file → | |
| IQ3_M | 53.8 GB (2 parts) | 55.8 GB | Too big | Get file → | |
| Q3_K_M | 53.8 GB (2 parts) | 55.9 GB | Too big | Get file → | |
| Q3_K_L | 54.2 GB (2 parts) | 56.3 GB | Too big | Get file → | |
| Q3_K_XL | 54.8 GB (2 parts) | 56.8 GB | Too big | Get file → | |
| IQ4_XS | 58.8 GB (2 parts) | 60.9 GB | Too big | Get file → | |
| IQ4_NL | 61.0 GB (2 parts) | 63.1 GB | Too big | Get file → | |
| Q4_0 | 61.7 GB (2 parts) | 63.8 GB | Too big | Get file → | |
| Q4_K_S | 66.2 GB (2 parts) | 68.2 GB | Too big | Get file → | |
| Q4_1 | 67.3 GB (2 parts) | 69.3 GB | Too big | Get file → | |
| Q4_K_M | 71.4 GB (2 parts) | 73.4 GB | Too big | Get file → | |
| Q4_K_L | 71.8 GB (2 parts) | 73.9 GB | Too big | Get file → | |
| Q5_K_S | 76.0 GB (2 parts) | 78.1 GB | Too big | Get file → | |
| Q5_K_M | 81.2 GB (3 parts) | 83.2 GB | Too big | Get file → | |
| Q6_K | 96.0 GB (3 parts) | 98.1 GB | Too big | Get file → | |
| Q8_0 | 113.6 GB (3 parts) | 115.7 GB | Too big | Get file → |
For 8 GB VRAM, use IQ1_S (35.29 GB file, 37.3 GB memory at 8 k context). For 16 GB VRAM, IQ2_XS (40.26 GB file, 42.3 GB memory). For 24 GB VRAM, Q3_K_S (51.74 GB file, 53.8 GB memory). The estimates are approximate.
What GPU do you need for INTELLECT 3?
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 (0.2 GB file, about 2.3 GB in use).
- 6 GB VRAM: the best quant that fits is the main file (0.2 GB file, about 2.3 GB in use).
- 8 GB VRAM: the best quant that fits is the main file (0.2 GB file, about 2.3 GB in use).
- 12 GB VRAM: the best quant that fits is the main file (0.2 GB file, about 2.3 GB in use).
- 16 GB VRAM: the best quant that fits is the main file (0.2 GB file, about 2.3 GB in use).
- 24 GB VRAM: the best quant that fits is the main file (0.2 GB file, about 2.3 GB in use).
- 32 GB VRAM: the best quant that fits is the main file (0.2 GB file, about 2.3 GB in use).
- 48 GB VRAM: the best quant that fits is Q2_K_L (45.3 GB file, about 47.4 GB in use).
Which INTELLECT 3 quant should you choose?
Q4_K_M is the usual starting point for INTELLECT 3: it balances file size and quality well. Here it is a 71.4 GB file.
The Q8_0 file is 113.6 GB, about 1.6 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 0.2 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 INTELLECT 3
How much VRAM does INTELLECT 3 need?
About 73.4 GB at Q4_K_M with an 8K context window. That is the 71.4 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 INTELLECT 3 on an 8 GB GPU?
Yes, with the smaller files: the 0.2 GB file needs about 2.3 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of INTELLECT 3 GGUF?
It depends on the quant: file is 0.2 GB, IQ1_S is 35.3 GB, IQ1_M is 35.9 GB, IQ2_XXS is 37.7 GB, IQ2_XS is 40.3 GB, IQ2_S is 40.6 GB and 21 more.
What context length does INTELLECT 3 support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 1.5 GB at 8K and 6.2 GB at 32K.
Where can I download INTELLECT 3 GGUF?
The files are on Hugging Face in the bartowski/PrimeIntellect_INTELLECT-3-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 INTELLECT 3 use?
Hugging Face lists the license as mit. 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 INTELLECT‑3 on 8 GB VRAM?
Use the IQ1_S quantization; it requires about 37.3 GB memory at 8 k context, which is beyond 8 GB VRAM, so it is not feasible.
Does INTELLECT‑3 support vision tasks?
No, the model has no vision capabilities.