Mistral-Small-4-119B-2603 needs about 74.6 GB of memory at Q4_K_M (a 73.8 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 23 quants, from 32.3 GB (IQ1_M) to 238.0 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 1.4 GB.
Mistral Small 4 119B – a multimodal, MoE LLM with 256k context
Mistral Small 4 is a 119 billion‑parameter model that supports text and image input. It uses a mixture‑of‑experts architecture with 128 experts, 4 active per token, and offers a 256k token context window. The model is licensed under Apache 2.0.
Good at
- Multimodal text‑image tasks
- Large‑context reasoning
- Function‑calling and JSON output
Limits
- Requires 8 k context for estimates
- Large VRAM needed for high‑bit quant
- No native vision support in llama.cpp without custom code
Best for
- Vision‑enabled chatbots
- Long‑form document analysis
- Agentic workflows with function calls
Who it’s for. Developers building multimodal applications that need large context and reasoning. The model is suitable for research and commercial use under Apache 2.0.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ1_M | 32.3 GB | 33.1 GB | Too big | Get file → | |
| IQ2_XXS | 34.9 GB | 35.8 GB | Too big | Get file → | |
| IQ2_M | 37.6 GB | 38.4 GB | Too big | Get file → | |
| Q2_K_XL | 40.2 GB | 41.0 GB | Too big | Get file → | |
| IQ3_XXS | 42.8 GB | 43.6 GB | Too big | Get file → | |
| IQ3_S | 44.4 GB | 45.2 GB | Too big | Get file → | |
| Q3_K_S | 49.6 GB | 50.5 GB | Too big | Get file → | |
| Q3_K_M | 54.4 GB (3 parts) | 55.2 GB | Too big | Get file → | |
| Q3_K_XL | 54.8 GB (3 parts) | 55.7 GB | Too big | Get file → | |
| IQ4_XS | 58.1 GB (3 parts) | 59.0 GB | Too big | Get file → | |
| IQ4_NL | 59.2 GB (3 parts) | 60.0 GB | Too big | Get file → | |
| Q4_K_S | 69.5 GB (3 parts) | 70.3 GB | Too big | Get file → | |
| MXFP4_MOE | 71.8 GB (3 parts) | 72.7 GB | Too big | Get file → | |
| Q4_K_M | 73.8 GB (3 parts) | 74.6 GB | Too big | Get file → | |
| Q4_K_XL | 74.2 GB (3 parts) | 75.0 GB | Too big | Get file → | |
| Q5_K_S | 83.6 GB (3 parts) | 84.4 GB | Too big | Get file → | |
| Q5_K_M | 89.2 GB (3 parts) | 90.0 GB | Too big | Get file → | |
| Q5_K_XL | 89.4 GB (3 parts) | 90.3 GB | Too big | Get file → | |
| Q6_K | 99.4 GB (4 parts) | 100.3 GB | Too big | Get file → | |
| Q6_K_XL | 108.3 GB (4 parts) | 109.1 GB | Too big | Get file → | |
| Q8_0 | 126.5 GB (4 parts) | 127.3 GB | Too big | Get file → | |
| Q8_K_XL | 132.6 GB (4 parts) | 133.4 GB | Too big | Get file → | |
| BF16 | 238.0 GB (6 parts) | 238.8 GB | Too big | Get file → |
For 8 GB VRAM, use IQ1_M (32.31 GB file, ~33 GB VRAM). With 16 GB VRAM, Q4_K_M (73.76 GB file, ~75 GB VRAM) is a good balance. For 32 GB VRAM, Q6_K_XL (108.27 GB file, ~109 GB VRAM) offers higher precision.
What GPU do you need for Mistral-Small-4-119B-2603?
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: no quant fits at an 8K context.
- 32 GB VRAM: no quant fits at an 8K context.
- 48 GB VRAM: the best quant that fits is IQ3_S (44.4 GB file, about 45.2 GB in use).
Which Mistral-Small-4-119B-2603 quant should you choose?
Q4_K_M is the usual starting point for Mistral-Small-4-119B-2603: it balances file size and quality well. Here it is a 73.8 GB file.
The Q8_0 file is 126.5 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 IQ1_M at 32.3 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Mistral-Small-4-119B-2603
How much VRAM does Mistral-Small-4-119B-2603 need?
About 74.6 GB at Q4_K_M with an 8K context window. That is the 73.8 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 Mistral-Small-4-119B-2603 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 33.1 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Mistral-Small-4-119B-2603 GGUF?
It depends on the quant: IQ1_M is 32.3 GB, IQ2_XXS is 34.9 GB, IQ2_M is 37.6 GB, Q2_K_XL is 40.2 GB, IQ3_XXS is 42.8 GB, IQ3_S is 44.4 GB and 17 more.
What context length does Mistral-Small-4-119B-2603 support?
The GGUF header lists a trained context length of 1,048,576 tokens (1024K). The KV cache grows with the context you use: about 0.3 GB at 8K and 1.4 GB at 32K.
Where can I download Mistral-Small-4-119B-2603 GGUF?
The files are on Hugging Face in the unsloth/Mistral-Small-4-119B-2603-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 Mistral-Small-4-119B-2603 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 Mistral Small 4 on 8 GB VRAM?
No, the smallest quant requires ~33 GB VRAM. 8 GB is insufficient for any quant of this model.
Does the model support vision in llama.cpp?
The base GGUF includes vision detection, but llama.cpp requires custom code or a patched build to handle image input.
What license does the model use?
Apache 2.0, allowing commercial and non‑commercial use with attribution.