Tawkeed-20B

Tawkeed's 20-billion parameter Mixture of Experts (MoE) model, purpose-built for Arabic language tasks. Delivers strong performance across Arabic text generation, conversational AI, summarization, and question answering.

Developed by: Tawkeed — Saudi-native enterprise Edge AI platform.

Use Cases

• Arabic text generation — produce fluent, coherent Arabic text across various domains
• Conversational AI — power Arabic chatbots and virtual assistants
• Summarization — condense long Arabic documents into concise summaries
• Question answering — answer questions using Arabic knowledge and context
• Content creation — draft articles, social media posts, and marketing copy in Arabic
• Translation assistance — aid in translating between Arabic and other languages

Usage

Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "tawkeed-sa/tawkeed-20b"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",
    torch_dtype="auto",
)

messages = [
    {"role": "system", "content": "أنت مساعد ذكي يتحدث العربية بطلاقة."},
    {"role": "user", "content": "ما هي أبرز إنجازات رؤية 2030 في المملكة العربية السعودية؟"},
]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to(model.device)
outputs = model.generate(inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0][inputs.shape[-1]:], skip_special_tokens=True))

vLLM

from vllm import LLM, SamplingParams

llm = LLM(model="tawkeed-sa/tawkeed-20b")
sampling_params = SamplingParams(temperature=0.7, max_tokens=512)

prompts = [
    "اكتب مقالاً قصيراً عن أهمية الذكاء الاصطناعي في التعليم",
    "لخص لي مفهوم الحوسبة السحابية في فقرة واحدة",
]
outputs = llm.generate(prompts, sampling_params)
for output in outputs:
    print(output.outputs[0].text)

OpenAI-Compatible API (vLLM)

# Start vLLM server
python -m vllm.entrypoints.openai.api_server --model tawkeed-sa/tawkeed-20b

# Query
curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tawkeed-sa/tawkeed-20b",
    "messages": [
      {"role": "system", "content": "أنت مساعد ذكي يتحدث العربية بطلاقة."},
      {"role": "user", "content": "اشرح لي مفهوم الذكاء الاصطناعي التوليدي"}
    ]
  }'

Model Details

Property	Value
Parameters	20B (Mixture of Experts)
Languages	Arabic, English
License	Apache 2.0
Context length	4096 tokens

About Tawkeed

Tawkeed is a Saudi-native enterprise Edge AI platform building Arabic-first AI models and tools. Our models cover 18 Arabic dialects and are optimized for on-premises deployment with complete data privacy.

All Tawkeed models  |  Arabic LLM Benchmark