TI

TinyMixtral 4x248M MoE i1

TinymixtralGGUF by mradermacherUpdated 2025-01-07
Download on Hugging Face
Parameters701M
Context32K
Layers12
Architecturellama
Smallest quant0.2 GB
Quick answer

TinyMixtral 4x248M MoE i1 needs about 1.0 GB of memory at Q4_K_M (a 0.4 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 24 quants, from 0.2 GB (IQ1_S) to 0.6 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 0.4 GB.

TinyMixtral 4x248M MoE i1 – a lightweight mixture‑of‑experts model

TinyMixtral 4x248M MoE i1 is a 0.701 billion‑parameter llama‑based model with a 32768‑token context window. It is a mixture of experts built from several TinyMistral variants. The model is released under the apache‑2.0 license.

Good at

  • Efficient inference with small quants
  • Large context handling
  • MoE architecture for specialized tasks

Limits

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

Best for

  • Low‑resource inference
  • Long‑form text generation
  • Fine‑tuning on small datasets

Who it’s for. Developers needing a compact MoE model for local deployment. Researchers can experiment with the mixture‑of‑experts design without large GPU memory.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S0.2 GB 0.8 GB
Fits Get file →
IQ1_M0.2 GB 0.8 GB
Fits Get file →
IQ2_XXS0.2 GB 0.8 GB
Fits Get file →
IQ2_XS0.2 GB 0.8 GB
Fits Get file →
IQ2_S0.2 GB 0.8 GB
Fits Get file →
IQ2_M0.2 GB 0.8 GB
Fits Get file →
Q2_K_S0.3 GB 0.9 GB
Fits Get file →
Q2_K0.3 GB 0.9 GB
Fits Get file →
IQ3_XXS0.3 GB 0.9 GB
Fits Get file →
IQ3_XS0.3 GB 0.9 GB
Fits Get file →
Q3_K_S0.3 GB 0.9 GB
Fits Get file →
IQ3_S0.3 GB 0.9 GB
Fits Get file →
IQ3_M0.3 GB 0.9 GB
Fits Get file →
Q3_K_M0.3 GB 0.9 GB
Fits Get file →
Q3_K_L0.4 GB 1.0 GB
Fits Get file →
IQ4_XS0.4 GB 1.0 GB
Fits Get file →
IQ4_NL0.4 GB 1.0 GB
Fits Get file →
Q4_00.4 GB 1.0 GB
Fits Get file →
Q4_K_S0.4 GB 1.0 GB
Fits Get file →
Q4_K_M0.4 GB 1.0 GB
Fits Get file →
Q4_10.4 GB 1.0 GB
Fits Get file →
Q5_K_S0.5 GB 1.1 GB
Fits Get file →
Q5_K_M0.5 GB 1.1 GB
Fits Get file →
Q6_K0.6 GB 1.2 GB
Fits Get file →

For 8 GB VRAM, IQ1_S or IQ1_M (0.16 GB or 0.18 GB) are suitable, using about 0.8 GB memory at 8 k context. If 16 GB VRAM is available, Q4_K_M (0.43 GB) or Q5_K_M (0.5 GB) provide higher quality while staying under 1.2 GB memory. Larger quants like Q6_K (0.58 GB) need more VRAM but offer better precision.

What GPU do you need for TinyMixtral 4x248M MoE i1?

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: the best quant that fits is Q6_K (0.6 GB file, about 1.2 GB in use).

Which TinyMixtral 4x248M MoE i1 quant should you choose?

Q4_K_M is the usual starting point for TinyMixtral 4x248M MoE i1: it balances file size and quality well. Here it is a 0.4 GB file.

If memory is tight, the smallest file is IQ1_S at 0.2 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
Parameters701M (approximate, mixture of experts)
Layers12
Trained context32,768 tokens
KV cache (fp16)12 KB per token

Source

Quantized bymradermacher
Licenseapache-2.0
Downloads (HF)1,033
Last updated2025-01-07

Frequently asked questions about TinyMixtral 4x248M MoE i1

How much VRAM does TinyMixtral 4x248M MoE i1 need?

About 1.0 GB at Q4_K_M with an 8K context window. That is the 0.4 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 TinyMixtral 4x248M MoE i1 on an 8 GB GPU?

Yes, with Q6_K: the 0.6 GB file needs about 1.2 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of TinyMixtral 4x248M MoE i1 GGUF?

It depends on the quant: IQ1_S is 0.2 GB, IQ1_M is 0.2 GB, IQ2_XXS is 0.2 GB, IQ2_XS is 0.2 GB, IQ2_S is 0.2 GB, IQ2_M is 0.2 GB and 18 more.

What context length does TinyMixtral 4x248M MoE i1 support?

The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 0.1 GB at 8K and 0.4 GB at 32K.

Where can I download TinyMixtral 4x248M MoE i1 GGUF?

The files are on Hugging Face in the mradermacher/TinyMixtral-4x248M-MoE-i1-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 TinyMixtral 4x248M MoE i1 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 TinyMixtral 4x248M MoE i1 on 8 GB VRAM?

Yes, using IQ1_S or IQ1_M quants will keep memory usage around 0.8 GB at 8 k context.

Does the model support vision tasks?

No, the model has no vision capabilities.

What license applies to this model?

The model is released under the apache‑2.0 license.

Other GGUFs of TinyMixtral-4x248M-MoE

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Tinymixtral models

All Tinymixtral →

Other tiny models

All tiny models →