MI

Ministral 3 3B Instruct 2512

MistralGGUF by ggml-orgUpdated 2025-12-02
Download on Hugging Face
Parameters3.43B
Context256K
Layers26
Architecturemistral3
Smallest quant3.7 GB
Quick answer

Ministral 3 3B Instruct 2512 needs about 5.0 GB of memory at Q8_0 (a 3.7 GB file) with an 8K context window, so it fits on a 6 GB GPU. Longer context needs more: at 32K tokens the KV cache alone adds about 3.5 GB.

Ministral 3 3B Instruct 2512 – a compact, vision‑enabled instruction model

Ministral 3 3B Instruct 2512 is a 3.43B parameter model with a 256k context window and integrated vision encoder. It is designed for edge deployment and supports multilingual instruction following.

Good at

  • Vision‑enabled text generation
  • Multilingual instruction following
  • Agentic function calling

Limits

  • Large context window requires 8k context for full memory estimate
  • No explicit vision support in non‑vision quantizations
  • Requires Apache 2.0 license compliance

Best for

  • Real‑time image captioning on edge devices
  • Multilingual chat assistants
  • Fine‑tuning for specialized tasks

Who it’s for. Developers building lightweight, real‑time AI applications on low‑resource hardware. Ideal for embedded systems needing vision and instruction capabilities without heavy GPU requirements.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q8_03.7 GB 5.0 GB
Fits Get file →

The only available quantization is Q8_0 with a file size of 3.65 GB. At an 8k context it estimates 5 GB of memory usage. This fits comfortably on systems with 8 GB VRAM or more. For tighter budgets, consider other formats listed in the full collection, but they are not provided here.

What GPU do you need for Ministral 3 3B 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: the best quant that fits is Q8_0 (3.7 GB file, about 5.0 GB in use).

Which Ministral 3 3B Instruct 2512 quant should you choose?

Q8_0 (3.7 GB) is the middle option in this repository. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturemistral3
Parameters3.43B
Layers26
Trained context262,144 tokens
KV cache (fp16)104 KB per token

Source

Quantized byggml-org
FamilyMistral
Licenseapache-2.0
Downloads (HF)1,912
Last updated2025-12-02

Frequently asked questions about Ministral 3 3B Instruct 2512

How much VRAM does Ministral 3 3B Instruct 2512 need?

About 5.0 GB at Q8_0 with an 8K context window. That is the 3.7 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 Ministral 3 3B Instruct 2512 on an 8 GB GPU?

Yes, with Q8_0: the 3.7 GB file needs about 5.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Ministral 3 3B Instruct 2512 GGUF?

It depends on the quant: Q8_0 is 3.7 GB.

What context length does Ministral 3 3B 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 0.9 GB at 8K and 3.5 GB at 32K.

Where can I download Ministral 3 3B Instruct 2512 GGUF?

The files are on Hugging Face in the ggml-org/Ministral-3-3B-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 Ministral 3 3B 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 this model on 8 GB VRAM?

Yes, the Q8_0 quantization requires an estimated 5 GB of VRAM at 8k context, which fits within 8 GB.

Does it support vision tasks?

Yes, the model includes a 0.4B vision encoder for image analysis and captioning.

What license applies?

The model is licensed under Apache 2.0, allowing commercial and non‑commercial use with attribution.

Other GGUFs of Ministral-3-3B-Instruct-2512-BF16

Similar models

Browse all →

More from ggml-org

All by ggml-org →

More Mistral models

All Mistral →

Other tiny models

All tiny models →