---

language:

• pt

license: gemma

library_name: gguf

pipeline_tag: text-generation

base_model: tardellirs/aac-board-generator-140m-ptbr

tags:

• gguf
• llama-cpp
• text-generation
• gemma3
• aac
• augmentative-alternative-communication
• arasaac
• board-generation
• brazilian-portuguese
• cpu

---

AAC Board Generator 140M (pt-BR) — GGUF

!Runs on




llama.cpp GGUF builds of aac-board-generator-140m-ptbr

— the ultra-light, ultra-fast Brazilian-Portuguese multi-function AAC model (141.3M, CPU): board generation +

in-assistant board-editing instructions. ⚠️ Lower quality than the

770M model; choose it for speed/footprint. See the

model card for details.

Files

| File | Quant | Size | Notes |

|---|---|---|---|

| aac-board-generator-140m-ptbr-Q4_K_M.gguf | Q4_K_M | 116 MB | smallest |

| aac-board-generator-140m-ptbr-Q5_K_M.gguf | Q5_K_M | 122 MB | balanced |

| aac-board-generator-140m-ptbr-Q6_K.gguf | Q6_K | 145 MB | near-lossless |

| aac-board-generator-140m-ptbr-Q8_0.gguf | Q8_0 | 154 MB | recommended — safe near-lossless |

| aac-board-generator-140m-ptbr-f16.gguf | f16 | 286 MB | full precision reference |

At Q8_0 the model serves in a few hundred MB of RAM and produces a ~12-item board in **under a second on 4 CPU

threads**.

Usage

llama-server -m aac-board-generator-140m-ptbr-Q8_0.gguf -t 4 -c 2048 --host 127.0.0.1 --port 8080

import requests
INSTR = ("Você monta pranchas de CAA (pictogramas, pt-BR). Para o PEDIDO, liste ~12 itens concretos e relevantes, "
         "um por linha, no formato palavra|tipo|sinônimos (tipo: v/s/a/e/l/p). Só a lista.")
pedido = "monta uma prancha de brincar no parquinho"
prompt = f"<start_of_turn>user\n{INSTR}\n\nPEDIDO: {pedido}<end_of_turn>\n<start_of_turn>model\n"
print(requests.post("http://127.0.0.1:8080/completion",
                    json={"prompt": prompt, "temperature": 0, "n_predict": 320}).json()["content"])

Greedy / deterministic (temperature 0); output is one item per line, word|type|synonyms.

Notes on the build

Like its 770M sibling, this is a vocabulary-trimmed Gemma 3 that uses the BPE tokenizer.json path (not a

reconstructed SentencePiece), so the GGUF tokenizes identically to the untrimmed parent and preserves greedy

generation.

License & attribution

Released under the Gemma license (inherited from Gemma 3). Built for the

ARASAAC ecosystem (CC BY-NC-SA content). See the

model card for full terms.

Developed for and used in Papuguinho — www.papuguinho.com.