MI

mixtral

MistralGGUF by hflUpdated 2024-03-27
Download on Hugging Face
Parameters46.7B
Context32K
Layers32
Architecturellama
Smallest quant9.8 GB
Quick answer

mixtral needs about 22.0 GB of memory at IQ3_S (a 20.4 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 20 quants, from 9.8 GB (IQ1_S) to 49.6 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.

Chinese-Mixtral-Instruct, 46.7B parameters, 32k context

Chinese-Mixtral-Instruct is a 46.7B parameter mixture‑of‑experts model based on the Mistral architecture. It supports 32,768 token context and is released under the Apache‑2.0 license. The model is available in multiple GGUF quantizations for llama.cpp.

Good at

  • Large context handling
  • Instruction‑driven Chinese dialogue
  • High‑capacity reasoning

Limits

  • Requires 8k context for memory estimates
  • No vision support
  • License is Apache‑2.0

Best for

  • Chatbot development
  • Content generation in Chinese
  • Research prototyping

Who it’s for. Developers and researchers needing a large Chinese instruction model that can run on llama.cpp. The model is suitable for local deployment with sufficient GPU memory.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S9.8 GB 11.4 GB
Too big Get file →
IQ1_M10.8 GB 12.4 GB
Too big Get file →
IQ2_XXS12.3 GB 13.9 GB
Too big Get file →
IQ2_XS13.7 GB 15.2 GB
Too big Get file →
IQ2_S14.1 GB 15.7 GB
Too big Get file →
IQ2_M15.5 GB 17.1 GB
Too big Get file →
Q2_K17.3 GB 18.9 GB
Too big Get file →
IQ3_XXS18.3 GB 19.9 GB
Too big Get file →
IQ3_XS19.3 GB 20.9 GB
Too big Get file →
IQ3_S20.4 GB 22.0 GB
Too big Get file →
IQ3_M21.4 GB 23.0 GB
Too big Get file →
Q3_K22.5 GB 24.1 GB
Too big Get file →
IQ4_XS25.1 GB 26.7 GB
Too big Get file →
Q4_026.4 GB 28.0 GB
Too big Get file →
IQ4_NL26.5 GB 28.1 GB
Too big Get file →
Q4_K28.4 GB 30.0 GB
Too big Get file →
Q5_032.2 GB 33.8 GB
Too big Get file →
Q5_K33.2 GB 34.8 GB
Too big Get file →
Q6_K38.4 GB 40.0 GB
Too big Get file →
Q8_049.6 GB 51.2 GB
Too big Get file →

For 8‑GB VRAM, use IQ1_S (9.82 GB, 11.4 GB memory at 8k context). For 12‑GB VRAM, IQ1_M (10.85 GB, 12.4 GB). For 16‑GB VRAM, IQ2_XS (13.67 GB, 15.2 GB). For 24‑GB VRAM, Q4_K (28.45 GB, 30 GB).

What GPU do you need for mixtral?

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: the best quant that fits is IQ1_S (9.8 GB file, about 11.4 GB in use).
  • 16 GB VRAM: the best quant that fits is IQ2_S (14.1 GB file, about 15.7 GB in use).
  • 24 GB VRAM: the best quant that fits is IQ3_M (21.4 GB file, about 23.0 GB in use).
  • 32 GB VRAM: the best quant that fits is Q4_K (28.4 GB file, about 30.0 GB in use).
  • 48 GB VRAM: the best quant that fits is Q6_K (38.4 GB file, about 40.0 GB in use).

Which mixtral quant should you choose?

IQ3_S (20.4 GB) is the middle option in this repository. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.

If memory is tight, the smallest file is IQ1_S at 9.8 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
Parameters46.7B (approximate, mixture of experts)
Layers32
Trained context32,768 tokens
KV cache (fp16)128 KB per token

Source

Quantized byhfl
FamilyMistral
Licenseapache-2.0
Downloads (HF)714
Last updated2024-03-27

Frequently asked questions about mixtral

How much VRAM does mixtral need?

About 22.0 GB at IQ3_S with an 8K context window. That is the 20.4 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 mixtral on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 11.4 GB, so you would need a larger GPU or CPU offloading.

What is the file size of mixtral GGUF?

It depends on the quant: IQ1_S is 9.8 GB, IQ1_M is 10.8 GB, IQ2_XXS is 12.3 GB, IQ2_XS is 13.7 GB, IQ2_S is 14.1 GB, IQ2_M is 15.5 GB and 14 more.

What context length does mixtral support?

The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 1.1 GB at 8K and 4.3 GB at 32K.

Where can I download mixtral GGUF?

The files are on Hugging Face in the hfl/chinese-mixtral-instruct-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 mixtral 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 this model on 8 GB VRAM?

Use the IQ1_S quantization; it requires about 11.4 GB memory at 8k context, which is close to 8 GB but may need CPU fallback.

Does the model support vision tasks?

No, the model is purely text‑based and has no vision capabilities.

What license applies to this model?

The model is released under the Apache‑2.0 license.

Similar models

Browse all →

More from hfl

All by hfl →

More Mistral models

All Mistral →

Other large models

All large models →