Devstral-2-123B-Instruct-2512 needs about 78.4 GB of memory at Q4_K_M (a 74.9 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 26 quants, from 28.4 GB (IQ1_S) to 250.1 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 11.8 GB.
Devstral‑2‑123B‑Instruct‑2512 – a large LLM for software engineering tasks
Devstral‑2‑123B‑Instruct‑2512 is a 123‑billion‑parameter instruction‑tuned model built on the Mistral architecture. It supports a 262,144‑token context window and is optimized for coding and agentic tasks. The model is available in multiple GGUF quantizations.
Good at
- Agentic coding and tool use
- Large context handling
- Instruction following
Limits
- Large VRAM requirement for higher quants
- No vision or multimodal support
- License is other – check usage terms
Best for
- AI code assistants
- Software engineering agents
- Complex codebase exploration
Who it’s for. Developers and teams needing a powerful coding assistant that can handle long documents and multiple files. The model is suitable for building agentic tools and integrating into development workflows.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ1_S | 28.4 GB | 31.9 GB | Too big | Get file → | |
| IQ1_M | 30.6 GB | 34.0 GB | Too big | Get file → | |
| IQ2_XXS | 34.4 GB | 37.9 GB | Too big | Get file → | |
| IQ2_M | 43.5 GB | 47.0 GB | Too big | Get file → | |
| Q2_K | 46.6 GB | 50.0 GB | Too big | Get file → | |
| Q2_K_L | 47.0 GB | 50.4 GB | Too big | Get file → | |
| Q2_K_XL | 47.9 GB | 51.4 GB | Too big | Get file → | |
| IQ3_XXS | 49.0 GB | 52.4 GB | Too big | Get file → | |
| Q3_K_S | 54.4 GB (2 parts) | 57.8 GB | Too big | Get file → | |
| Q3_K_M | 60.6 GB (2 parts) | 64.1 GB | Too big | Get file → | |
| Q3_K_XL | 61.9 GB (2 parts) | 65.4 GB | Too big | Get file → | |
| IQ4_XS | 67.1 GB (2 parts) | 70.6 GB | Too big | Get file → | |
| IQ4_NL | 70.9 GB (2 parts) | 74.3 GB | Too big | Get file → | |
| Q4_0 | 71.0 GB (2 parts) | 74.5 GB | Too big | Get file → | |
| Q4_K_S | 71.2 GB (2 parts) | 74.7 GB | Too big | Get file → | |
| Q4_K_M | 74.9 GB (2 parts) | 78.4 GB | Too big | Get file → | |
| Q4_K_XL | 75.5 GB (2 parts) | 79.0 GB | Too big | Get file → | |
| Q4_1 | 78.5 GB (2 parts) | 81.9 GB | Too big | Get file → | |
| Q5_K_S | 86.2 GB (2 parts) | 89.6 GB | Too big | Get file → | |
| Q5_K_XL | 88.2 GB (2 parts) | 91.7 GB | Too big | Get file → | |
| Q5_K_M | 88.3 GB (2 parts) | 91.8 GB | Too big | Get file → | |
| Q6_K | 102.6 GB (3 parts) | 106.0 GB | Too big | Get file → | |
| Q6_K_XL | 109.8 GB (3 parts) | 113.3 GB | Too big | Get file → | |
| Q8_0 | 132.9 GB (3 parts) | 136.3 GB | Too big | Get file → | |
| Q8_K_XL | 146.1 GB (3 parts) | 149.5 GB | Too big | Get file → | |
| BF16 | 250.1 GB (6 parts) | 253.5 GB | Too big | Get file → |
For 8 GB VRAM, use IQ1_S (28.45 GB file, ~32 GB VRAM). With 12 GB VRAM, IQ1_M (30.57 GB, ~34 GB VRAM) is suitable. For 16 GB VRAM, Q2_K (46.59 GB, ~50 GB VRAM) or Q2_K_L (46.97 GB, ~50.4 GB VRAM) can be used. Higher quants require more memory.
What GPU do you need for Devstral-2-123B-Instruct-2512?
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: the best quant that fits is IQ1_S (28.4 GB file, about 31.9 GB in use).
- 48 GB VRAM: the best quant that fits is IQ2_M (43.5 GB file, about 47.0 GB in use).
Which Devstral-2-123B-Instruct-2512 quant should you choose?
Q4_K_M is the usual starting point for Devstral-2-123B-Instruct-2512: it balances file size and quality well. Here it is a 74.9 GB file.
The Q8_0 file is 132.9 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 IQ1_S at 28.4 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 Devstral-2-123B-Instruct-2512
How much VRAM does Devstral-2-123B-Instruct-2512 need?
About 78.4 GB at Q4_K_M with an 8K context window. That is the 74.9 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 Devstral-2-123B-Instruct-2512 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 31.9 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Devstral-2-123B-Instruct-2512 GGUF?
It depends on the quant: IQ1_S is 28.4 GB, IQ1_M is 30.6 GB, IQ2_XXS is 34.4 GB, IQ2_M is 43.5 GB, Q2_K is 46.6 GB, Q2_K_L is 47.0 GB and 20 more.
What context length does Devstral-2-123B-Instruct-2512 support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 3.0 GB at 8K and 11.8 GB at 32K.
Where can I download Devstral-2-123B-Instruct-2512 GGUF?
The files are on Hugging Face in the unsloth/Devstral-2-123B-Instruct-2512-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 Devstral-2-123B-Instruct-2512 use?
Hugging Face lists the license as other. 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 Devstral‑2‑123B‑Instruct‑2512 on 8 GB VRAM?
Use the IQ1_S quant, which requires about 32 GB VRAM at 8k context. 8 GB VRAM is insufficient for any quant of this model.
Does the model support vision or multimodal inputs?
No, the model only processes text. It has no vision or multimodal capabilities.
What context length does the model support?
The model supports a 262,144‑token context window, allowing very long documents to be processed.