Voho Saudi Speak
0.6B

chrF++ against what a Saudi speaker actually said, on 1,000 held-out sentences (higher is better).

"Formal text unchanged" is the floor: dialect and formal Arabic share most of their letters, so leaving a sentence alone already scores well. Gemini 2.5 Flash, a much larger model given the same instruction, is shown for scale. This model learned the test data’s style of Saudi spelling and Gemini did not, so the comparison favours the small model on this test; it does not mean it is better at Saudi Arabic in general.

Run it locally with Ollama

ollama run hf.co/VohoAI/voho-saudi-speak-0.6b-GGUF:Q4_K_M

Python, with Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "VohoAI/voho-saudi-speak-0.6b"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(repo)

DIALECT = {"najdi": "النجدية", "hijazi": "الحجازية", "khaleeji": "الخليجية الشرقية"}

def saudi(text, dialect="najdi"):
    prompt = f"أعد صياغة هذه الجملة باللهجة السعودية {DIALECT[dialect]} كما يقولها شخص في مكالمة، بدون أي شرح:\n{text}"
    ids = tok.apply_chat_template([{"role": "user", "content": prompt}], add_generation_prompt=True,
                                  enable_thinking=False, return_tensors="pt")
    out = model.generate(ids, max_new_tokens=96, do_sample=False)
    return tok.decode(out[0, ids.shape[1]:], skip_special_tokens=True).strip()

print(saudi("أين أنت الآن؟ أريد أن أحجز موعداً غداً."))