MI

Mistral-Large-3-675B-Instruct-2512

MistralGGUF by unslothUpdated 2025-12-16
Download on Hugging Face
Parameters675B
Context288K
Layers61
Architecturedeepseek2
Smallest quant168.6 GB
Quick answer

Mistral-Large-3-675B-Instruct-2512 needs about 408.6 GB of memory at Q4_K_M (a 407.0 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 27 quants, from 168.6 GB (TQ1_0) to 1347.0 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.

Mistral‑Large‑3‑675B Instruct 2512 – a multimodal MoE model with 675B parameters

Mistral‑Large‑3‑675B Instruct 2512 is a multimodal granular Mixture‑of‑Experts model with 675B total parameters and 41B active parameters. It supports vision, instruction following, and a 256k context window. The model is released under Apache‑2.0.

Good at

  • Vision analysis
  • Long‑context instruction following
  • Agentic function calling

Limits

  • Large VRAM requirement – 170.2 GB at 8k context (TQ1_0)
  • No GPU‑specific optimisations listed
  • Requires 256k context window for full capability

Best for

  • Long document understanding
  • Enterprise knowledge work
  • Agentic tool‑use

Who it’s for. Developers building production‑grade assistants, retrieval‑augmented systems, or enterprise workflows that need multimodal input and long‑context reasoning. The model is suitable for commercial use under Apache‑2.0.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
TQ1_0168.6 GB 170.2 GB
Too big Get file →
IQ1_S190.8 GB (4 parts) 192.4 GB
Too big Get file →
IQ1_M206.5 GB (5 parts) 208.0 GB
Too big Get file →
IQ2_XXS224.8 GB (5 parts) 226.4 GB
Too big Get file →
IQ2_M235.4 GB (5 parts) 237.0 GB
Too big Get file →
Q2_K247.1 GB (5 parts) 248.7 GB
Too big Get file →
Q2_K_L247.3 GB (6 parts) 248.9 GB
Too big Get file →
Q2_K_XL256.4 GB (6 parts) 258.0 GB
Too big Get file →
IQ3_XXS279.2 GB (6 parts) 280.8 GB
Too big Get file →
Q3_K_S291.9 GB (6 parts) 293.4 GB
Too big Get file →
Q3_K_XL300.8 GB (7 parts) 302.4 GB
Too big Get file →
Q3_K_M321.8 GB (7 parts) 323.4 GB
Too big Get file →
IQ4_XS360.0 GB (8 parts) 361.5 GB
Too big Get file →
IQ4_NL380.7 GB (8 parts) 382.3 GB
Too big Get file →
Q4_0381.7 GB (8 parts) 383.3 GB
Too big Get file →
Q4_K_S382.7 GB (8 parts) 384.2 GB
Too big Get file →
Q4_K_XL387.9 GB (8 parts) 389.5 GB
Too big Get file →
Q4_K_M407.0 GB (9 parts) 408.6 GB
Too big Get file →
Q4_1422.4 GB (9 parts) 424.0 GB
Too big Get file →
Q5_K_S464.4 GB (10 parts) 466.0 GB
Too big Get file →
Q5_K_M478.0 GB (10 parts) 479.6 GB
Too big Get file →
Q5_K_XL479.8 GB (10 parts) 481.4 GB
Too big Get file →
Q6_K553.3 GB (12 parts) 554.9 GB
Too big Get file →
Q6_K_XL576.5 GB (13 parts) 578.1 GB
Too big Get file →
Q8_0715.7 GB (15 parts) 717.3 GB
Too big Get file →
Q8_K_XL773.5 GB (17 parts) 775.1 GB
Too big Get file →
BF161347.0 GB (30 parts) 1348.5 GB
Too big Get file →

For 8 GB VRAM use Q4_0 (381.7 GB file, 383.3 GB memory at 8k). For 16 GB VRAM use Q4_K_S (382.66 GB file, 384.2 GB memory). For 32 GB VRAM use Q5_K_S (464.45 GB file, 466 GB memory).

What GPU do you need for Mistral-Large-3-675B-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: no quant fits at an 8K context.
  • 48 GB VRAM: no quant fits at an 8K context.

Which Mistral-Large-3-675B-Instruct-2512 quant should you choose?

Q4_K_M is the usual starting point for Mistral-Large-3-675B-Instruct-2512: it balances file size and quality well. Here it is a 407.0 GB file.

The Q8_0 file is 715.7 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 TQ1_0 at 168.6 GB, with the quality trade-offs that come with fewer bits. Lower bit counts mean smaller files and more quality loss.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturedeepseek2
Parameters675B (approximate, mixture of experts)
Layers61
Trained context294,912 tokens
KV cache (fp16)130 KB per token

Source

Quantized byunsloth
FamilyMistral
Licenseapache-2.0
Downloads (HF)2,972
Last updated2025-12-16

Frequently asked questions about Mistral-Large-3-675B-Instruct-2512

How much VRAM does Mistral-Large-3-675B-Instruct-2512 need?

About 408.6 GB at Q4_K_M with an 8K context window. That is the 407.0 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-3-675B-Instruct-2512 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 170.2 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Mistral-Large-3-675B-Instruct-2512 GGUF?

It depends on the quant: TQ1_0 is 168.6 GB, IQ1_S is 190.8 GB, IQ1_M is 206.5 GB, IQ2_XXS is 224.8 GB, IQ2_M is 235.4 GB, Q2_K is 247.1 GB and 21 more.

What context length does Mistral-Large-3-675B-Instruct-2512 support?

The GGUF header lists a trained context length of 294,912 tokens (288K). 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 Mistral-Large-3-675B-Instruct-2512 GGUF?

The files are on Hugging Face in the unsloth/Mistral-Large-3-675B-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 Mistral-Large-3-675B-Instruct-2512 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‑Large‑3 on 8 GB VRAM?

No, the smallest quant requires 170.2 GB memory at 8k context. 8 GB VRAM is insufficient.

Does the model support vision?

Yes, it includes a 2.5B vision encoder for image analysis.

What license applies?

The model is released under the Apache‑2.0 license, allowing commercial use.

Other GGUFs of Mistral-Large-3-675B-Instruct-2512

Similar models

Browse all →

More from unsloth

All by unsloth →

More Mistral models

All Mistral →

Other large models

All large models →