MI

Mistral Large Instruct

MistralGGUF by InferenceIllusionistUpdated 2024-08-04
Download on Hugging Face
Parameters123B
Context32K
Layers88
Architecturellama
Smallest quant26.0 GB
Quick answer

Mistral Large Instruct needs about 76.7 GB of memory at Q4_K_M (a 73.2 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 20 quants, from 26.0 GB (IQ1_S) to 130.3 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 11.8 GB.

Mistral Large Instruct 123B quantized for local inference

Mistral Large Instruct 123B is a llama architecture model quantized by InferenceIllusionist. It supports a 32768 token context window and is available in multiple quantizations ranging from IQ1_S to Q8_0. The model is released under a non‑standard license.

Good at

  • Long‑form text generation
  • Instruction following
  • Large context handling

Limits

  • No vision support
  • License not standard
  • Requires large VRAM for higher quants

Best for

  • Chatbot development
  • Content drafting
  • Research prototyping

Who it’s for. Developers and researchers who need a large‑parameter instruction model for local deployment. The model is suitable for those with access to GPUs capable of handling 8‑k context inference.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S26.0 GB 29.4 GB
Too big Get file →
IQ1_M28.4 GB 31.8 GB
Too big Get file →
IQ2_XXS32.4 GB 35.9 GB
Too big Get file →
IQ2_XS36.1 GB 39.5 GB
Too big Get file →
IQ2_S38.4 GB 41.8 GB
Too big Get file →
IQ2_M41.6 GB 45.1 GB
Too big Get file →
Q2_K45.2 GB 48.6 GB
Too big Get file →
IQ3_XXS47.0 GB 50.5 GB
Too big Get file →
IQ3_XS50.1 GB (2 parts) 53.6 GB
Too big Get file →
Q3_K_S52.8 GB (2 parts) 56.3 GB
Too big Get file →
IQ3_S53.0 GB (2 parts) 56.4 GB
Too big Get file →
IQ3_M55.3 GB (2 parts) 58.7 GB
Too big Get file →
Q3_K_M59.1 GB (2 parts) 62.6 GB
Too big Get file →
IQ4_XS65.4 GB (2 parts) 68.9 GB
Too big Get file →
Q4_K_S69.6 GB (2 parts) 73.0 GB
Too big Get file →
Q4_K_M73.2 GB (2 parts) 76.7 GB
Too big Get file →
Q5_K_S84.4 GB (2 parts) 87.8 GB
Too big Get file →
Q5_K_M86.5 GB (2 parts) 89.9 GB
Too big Get file →
Q6_K100.6 GB (3 parts) 104.0 GB
Too big Get file →
Q8_0130.3 GB (3 parts) 133.7 GB
Too big Get file →

For 8‑GB VRAM use IQ1_S (25.96 GB, 29.4 GB memory at 8k). For 16‑GB VRAM use IQ2_M (41.62 GB, 45.1 GB memory). For 24‑GB VRAM use Q3_K_M (59.1 GB, 62.6 GB memory). Estimates are approximate.

What GPU do you need for Mistral Large Instruct?

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_M (28.4 GB file, about 31.8 GB in use).
  • 48 GB VRAM: the best quant that fits is IQ2_M (41.6 GB file, about 45.1 GB in use).

Which Mistral Large Instruct quant should you choose?

Q4_K_M is the usual starting point for Mistral Large Instruct: it balances file size and quality well. Here it is a 73.2 GB file.

The Q8_0 file is 130.3 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 26.0 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

Architecturellama
Parameters123B
Layers88
Trained context32,768 tokens
KV cache (fp16)352 KB per token

Source

FamilyMistral
Licenseother
Downloads (HF)630
Last updated2024-08-04

Frequently asked questions about Mistral Large Instruct

How much VRAM does Mistral Large Instruct need?

About 76.7 GB at Q4_K_M with an 8K context window. That is the 73.2 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 Large Instruct on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 29.4 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Mistral Large Instruct GGUF?

It depends on the quant: IQ1_S is 26.0 GB, IQ1_M is 28.4 GB, IQ2_XXS is 32.4 GB, IQ2_XS is 36.1 GB, IQ2_S is 38.4 GB, IQ2_M is 41.6 GB and 14 more.

What context length does Mistral Large Instruct support?

The GGUF header lists a trained context length of 32,768 tokens (32K). 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 Mistral Large Instruct GGUF?

The files are on Hugging Face in the InferenceIllusionist/Mistral-Large-Instruct-2407-iMat-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 Large Instruct 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 Mistral Large Instruct on 8 GB VRAM?

Use the IQ1_S quantization; it requires about 29.4 GB memory at 8k context, which is above 8 GB, so it is not feasible on 8 GB VRAM.

Does this model support vision tasks?

No, the model has no vision capabilities.

What context length does it support?

It supports a 32768 token context window.

Other GGUFs of Mistral-Large-Instruct-2407

Similar models

Browse all →

More from InferenceIllusionist

All by InferenceIllusionist →

More Mistral models

All Mistral →

Other large models

All large models →