GE

gemmuguinho 636m

GemmuguinhoGGUF by PapuguinhoUpdated 2026-10-20
Download on Hugging Face
Parameters636M
Context32K
Layers18
Architecturegemma3moe
Smallest quant0.5 GB
Quick answer

gemmuguinho 636m needs about 1.1 GB of memory at Q4_K_M (a 0.5 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 10 quants, from 0.5 GB (Q4_K_M) to 1.3 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 0.6 GB.

Compact MoE model for AAC pictogram boards in Brazilian Portuguese

Gemmuguinho 636m is a 636M‑parameter Mixture‑of‑Experts model built on gemma3moe. It generates and edits AAC pictogram boards on CPU via llama.cpp. The model runs fully on-device with no GPU.

Good at

  • AAC board generation
  • Board editing via natural‑language commands
  • Runs on CPU, no GPU needed

Limits

  • Requires custom gemma3moe architecture in llama.cpp
  • Limited to Brazilian Portuguese
  • Small context size (32768 tokens) but high memory for larger quants

Best for

  • Speech‑therapy AAC board creation
  • Assistive communication tool for non‑verbal users
  • Rapid board editing on commodity hardware

Who it’s for. Developers and clinicians building AAC tools for Brazilian Portuguese users. Ideal for on‑device applications where GPU resources are unavailable.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q4_K_M0.5 GB 1.1 GB
Fits Get file →
Q4_K_M0.5 GB 1.1 GB
Fits Get file →
Q5_K_M0.5 GB 1.1 GB
Fits Get file →
Q5_K_M0.5 GB 1.1 GB
Fits Get file →
Q6_K0.6 GB 1.3 GB
Fits Get file →
Q6_K0.6 GB 1.3 GB
Fits Get file →
Q8_00.7 GB 1.3 GB
Fits Get file →
Q8_00.7 GB 1.3 GB
Fits Get file →
F161.3 GB 1.9 GB
Fits Get file →
F161.3 GB 1.9 GB
Fits Get file →

For 8 GB VRAM, choose Q4_K_M (460 MB, ~1.1 GB RAM at 8 k context). If more RAM is available, Q5_K_M (500 MB) offers better quality. For best fidelity, use Q8_0 (680 MB) but it needs ~1.3 GB RAM at 8 k context.

What GPU do you need for gemmuguinho 636m?

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 F16 (1.3 GB file, about 1.9 GB in use).

Which gemmuguinho 636m quant should you choose?

Q4_K_M is the usual starting point for gemmuguinho 636m: it balances file size and quality well. Here it is a 0.5 GB file.

The Q8_0 file is 0.7 GB, about 1.5 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.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturegemma3moe
Parameters636M (approximate, mixture of experts)
Layers18
Trained context32,768 tokens
KV cache (fp16)18 KB per token

Source

Quantized byPapuguinho
Licensegemma
Downloads (HF)3,307
Last updated2026-10-20

Frequently asked questions about gemmuguinho 636m

How much VRAM does gemmuguinho 636m need?

About 1.1 GB at Q4_K_M with an 8K context window. That is the 0.5 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 gemmuguinho 636m on an 8 GB GPU?

Yes, with F16: the 1.3 GB file needs about 1.9 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of gemmuguinho 636m GGUF?

It depends on the quant: Q4_K_M is 0.5 GB, Q4_K_M is 0.5 GB, Q5_K_M is 0.5 GB, Q5_K_M is 0.5 GB, Q6_K is 0.6 GB, Q6_K is 0.6 GB and 4 more.

What context length does gemmuguinho 636m support?

The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 0.2 GB at 8K and 0.6 GB at 32K.

Where can I download gemmuguinho 636m GGUF?

The files are on Hugging Face in the Papuguinho/gemmuguinho-636m-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 gemmuguinho 636m use?

Hugging Face lists the license as gemma. 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 Gemmuguinho 636m on a laptop with 8 GB VRAM?

Yes, the Q4_K_M quant uses about 1.1 GB RAM at 8 k context, fitting comfortably on 8 GB VRAM systems.

Does the model support English?

No, the model is trained for Brazilian Portuguese only.

Do I need a GPU to run this model?

No, it runs fully on CPU via llama.cpp with a custom gemma3moe architecture.

Similar models

Browse all →

More from Papuguinho

All by Papuguinho →

Other tiny models

All tiny models →