MuchKnow-Swahili-8B 🇹🇿🇬🇧

MuchKnow-Swahili-8B is a high-performance bilingual (Tanzanian Kiswahili 🇹🇿 + English 🇬🇧) Large Language Model fine-tuned on top of deepseek-ai/DeepSeek-R1-Distill-Llama-8B.

Developed for MuchKnow (muchknow.com) and copyrighted to BuruOps (buruops.com), this model understands and generates authentic East African Kiswahili across three distinct linguistic registers alongside fluent English.

---

🌟 Key Features

1. Triple Kiswahili Register Alignment (Sajili Tatu za Kiswahili):
   • Kiswahili Sanifu (Standard Swahili): Formal communication, official correspondence, and educational contexts based on BAKITA standards.
   • Kiswahili Fasaha (Refined / Literary Swahili): Sophisticated, grammatically rich, and expressive Swahili suitable for literature, speeches, and deep technical discourse.
   • Kiswahili cha Mtaani (Street Swahili / Sheng): Authentic urban street vernacular used across Tanzanian cities (Dar es Salaam, Arusha, Mwanza, Dodoma) for natural conversational engagement.
2. Native English & Kiswahili Code-Switching:
   • Seamlessly comprehends queries in English, Kiswahili, or mixed bilingual inputs, and responds in the user's preferred language.
3. Advanced Reasoning:
   • Inherits DeepSeek-R1's strong mathematical, logic, and reasoning capabilities distilled into an efficient 8B parameter model.

---

🚀 Quickstart Usage

Using mlx_lm (Apple Silicon Mac)

from mlx_lm import load, generate

model, tokenizer = load("Bur3hani/MuchKnow-Swahili-8B")

prompt = """Below is an instruction that describes a task. Write a response that appropriately completes the request.

### Instruction:
Eleza kwa Kiswahili Sanifu na Kiswahili cha Mtaani jinsi ya kuanzisha biashara ya mtandaoni nchini Tanzania.

### Response:
"""

response = generate(model, tokenizer, prompt=prompt, max_tokens=512)
print(response)

---

📜 Copyright & License

• Copyright: © 2026 BuruOps (buruops.com) & MuchKnow (muchknow.com). All Rights Reserved.
• Base Model License: [Apache 2.0 / MIT] derived from DeepSeek-R1-Distill-Llama-8B.