Instructions to use engomeryasir/Jarvis-Qwen3-8B-Turkish-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use engomeryasir/Jarvis-Qwen3-8B-Turkish-v2 with MLX:

# Make sure mlx-lm is installed
# pip install --upgrade mlx-lm

# Generate text with mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("engomeryasir/Jarvis-Qwen3-8B-Turkish-v2")

prompt = "Write a story about Einstein"
messages = [{"role": "user", "content": prompt}]
prompt = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True
)

text = generate(model, tokenizer, prompt=prompt, verbose=True)

Notebooks
Google Colab
Kaggle
Local Apps Settings
LM Studio

How to use engomeryasir/Jarvis-Qwen3-8B-Turkish-v2 with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "engomeryasir/Jarvis-Qwen3-8B-Turkish-v2"

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "engomeryasir/Jarvis-Qwen3-8B-Turkish-v2"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use engomeryasir/Jarvis-Qwen3-8B-Turkish-v2 with Hermes Agent:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "engomeryasir/Jarvis-Qwen3-8B-Turkish-v2"

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default engomeryasir/Jarvis-Qwen3-8B-Turkish-v2

Run Hermes

hermes

MLX LM

How to use engomeryasir/Jarvis-Qwen3-8B-Turkish-v2 with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "engomeryasir/Jarvis-Qwen3-8B-Turkish-v2"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "engomeryasir/Jarvis-Qwen3-8B-Turkish-v2"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "engomeryasir/Jarvis-Qwen3-8B-Turkish-v2",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Jarvis-Qwen3-8B-Turkish-v2

Türkçe kod yazma, reasoning ve tool calling yeteneklerine sahip fine-tuned Qwen3-8B modeli.

A fine-tuned Qwen3-8B model specialized for Turkish code generation, chain-of-thought reasoning, and tool calling capabilities. Trained on Apple Silicon using MLX framework with QLoRA.

Model Özeti / Model Summary

Özellik	Değer
Base Model	Qwen/Qwen3-8B
Parameters	8.19B (6.06M trainable via LoRA)
Quantization	Merged QLoRA adapters (rank=16, 10 layers)
Framework	MLX (Apple Silicon optimized)
Training Hardware	Apple M4 Pro 24GB Unified Memory
License	Apache-2.0
Language	Turkish (tr), English (en)

Eğitim Detayları / Training Details

3-Aşamalı Curriculum Pipeline

Model, 3 aşamalı bir curriculum learning pipeline ile eğitilmiştir:

Aşama	İterasyon	Learning Rate	Açıklama
SFT-General	800	2e-5	Genel talimat takibi + temel kod yazma (zorluk 1-3)
SFT-Expert	500	1e-5	İleri kod + algoritma + sistem tasarımı (zorluk 3-5)
DPO Alignment	200	5e-6	Kod kalitesi, verimlilik, güvenlik hizalaması

Eğitim Verisi

Toplam 1701 filtreli örnek (2249 ham → %24.4 azalma, diversity score: 0.877):

Kategori	Örnek Sayısı	Metodoloji
Kod (14 kategori)	780	Evol-Instruct + OSS-Instruct + SelfCodeAlign
Chain-of-Thought	580	GPT-4o-mini ile 6 reasoning kategorisi
Tool Calling	245	Hermes function calling format
DPO Pairs	214	Kod kalitesi, verimlilik, güvenlik, test, açıklama
Türkçe Kültür	160	Atasözleri, dilbilgisi, tarih, internet jargonu

Kod kategorileri: Python (fundamentals + advanced), algorithms & data structures, data science, web development, JavaScript/TypeScript, SQL, DevOps/Shell, testing, design patterns, agent development, system design, debugging, Java enterprise.

Veri Üretim Yöntemleri

Evol-Instruct (WizardCoder tarzı): Basit sorudan başlayıp iteratif karmaşıklaştırma
OSS-Instruct (Magicoder tarzı): Seed örneklerden türetme
SelfCodeAlign: Sandbox doğrulama ile kod kalite kontrolü
AlpaGasus + DeITA: Hibrit LLM kalite filtreleme
Embedding Deduplication: Semantik tekrar eliminasyonu

Benchmark Sonuçları / Evaluation Results

Benchmark	Skor	Açıklama
HumanEval-TR	70% pass@1	Türkçe açıklamalı klasik HumanEval problemleri
MBPP-TR	50% pass@1	Basit Python görevleri, 3 test case ile
CodeQA-TR	100%	Kod açıklama ve hata bulma soruları
MultiLang	100%	JavaScript (%100), SQL (%100)
TR-BFCL	37.5%	Türkçe function calling (multi-tool %100, negative %100)

Kullanım / Usage

MLX ile (Apple Silicon)

from mlx_lm import load, generate

model, tokenizer = load("engomeryasir/Jarvis-Qwen3-8B-Turkish-v2")

prompt = "Python ile binary search algoritması yaz ve karmaşıklığını açıkla."
messages = [{"role": "user", "content": prompt}]
text = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)

response = generate(model, tokenizer, prompt=text, max_tokens=1024)
print(response)

MLX Server ile

# Sunucuyu başlat
python -m mlx_lm.server --model engomeryasir/Jarvis-Qwen3-8B-Turkish-v2 --port 11435

# API isteği
curl http://localhost:11435/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"messages":[{"role":"user","content":"FastAPI ile REST API yaz"}],"max_tokens":500}'

Özellikler / Features

Türkçe Kod Yazma: 14 programlama kategorisinde Türkçe talimatlarla kod üretimi
Chain-of-Thought: <think>...</think> blokları ile şeffaf düşünme süreci
Tool Calling: Hermes function calling formatında araç kullanımı
Çok Dilli Kod: Python, JavaScript, TypeScript, SQL, Bash, Java
Türkçe Kültür: Atasözleri, dilbilgisi, tarih bilgisi

Sınırlamalar / Limitations

Model ağırlıklı olarak İngilizce pre-training'e sahiptir; Türkçe performansı sınırlı olabilir
Tool calling formatı Hermes standardındadır; diğer formatlarla uyumsuz olabilir
4.3GB model boyutu — minimum 8GB RAM önerilir
Bash kod üretiminde düşük performans (%0)
Hallucination riski mevcuttur; kritik uygulamalarda doğrulama önerilir

Alıntı / Citation

@misc{jarvis-qwen3-8b-turkish-v2,
  title={Jarvis-Qwen3-8B-Turkish-v2: A Fine-tuned Turkish Code & Reasoning Model},
  author={Ömer Yasir Önal},
  year={2026},
  url={https://huggingface.co/engomeryasir/Jarvis-Qwen3-8B-Turkish-v2}
}

Teşekkürler / Acknowledgments

Qwen Team — Base model
MLX Team — Apple Silicon ML framework
WizardCoder — Evol-Instruct methodology
Magicoder — OSS-Instruct methodology
SelfCodeAlign — Self-play alignment for code

Downloads last month: 17

Safetensors

Model size

1B params

Tensor type

BF16

U32

MLX

Hardware compatibility

4-bit

Model tree for engomeryasir/Jarvis-Qwen3-8B-Turkish-v2

Base model

Qwen/Qwen3-8B-Base

Finetuned

Qwen/Qwen3-8B

Quantized

(291)

this model

Papers for engomeryasir/Jarvis-Qwen3-8B-Turkish-v2

Evaluation results

pass@1
self-reported

0.700
pass@1
self-reported

0.500
accuracy
self-reported

1.000
accuracy
self-reported

0.375