Instructions to use ecloudtech/Erk-14B-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use ecloudtech/Erk-14B-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf ecloudtech/Erk-14B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf ecloudtech/Erk-14B-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf ecloudtech/Erk-14B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf ecloudtech/Erk-14B-GGUF:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf ecloudtech/Erk-14B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf ecloudtech/Erk-14B-GGUF:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf ecloudtech/Erk-14B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf ecloudtech/Erk-14B-GGUF:Q4_K_M

Use Docker

docker model run hf.co/ecloudtech/Erk-14B-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use ecloudtech/Erk-14B-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ecloudtech/Erk-14B-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ecloudtech/Erk-14B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ecloudtech/Erk-14B-GGUF:Q4_K_M

Ollama
How to use ecloudtech/Erk-14B-GGUF with Ollama:
```
ollama run hf.co/ecloudtech/Erk-14B-GGUF:Q4_K_M
```
Unsloth Desktop

How to use ecloudtech/Erk-14B-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ecloudtech/Erk-14B-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "ecloudtech/Erk-14B-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use ecloudtech/Erk-14B-GGUF with Docker Model Runner:
```
docker model run hf.co/ecloudtech/Erk-14B-GGUF:Q4_K_M
```

Lemonade

How to use ecloudtech/Erk-14B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull ecloudtech/Erk-14B-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Erk-14B-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use ecloudtech/Erk-14B-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ecloudtech/Erk-14B-GGUF:Q4_K_M

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 ecloudtech/Erk-14B-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use ecloudtech/Erk-14B-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ecloudtech/Erk-14B-GGUF:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "ecloudtech/Erk-14B-GGUF:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Erk-14B — GGUF

ecloudtech/Erk-14B modelinin llama.cpp biçimindeki sürümleri. eCloud Tech. tarafından üretildi ve her sürüm ayrı ayrı sınandı — nicemleme sonrası davranış ölçülmeden yayımlanmaz.

English below.

Sürümler

Dosya	Boyut	Kimlik sınavı	NLL	Kimin için
`Erk-14B-BF16.gguf`	29,5 GB	6/6	1,5304	kayıpsız referans
`Erk-14B-Q4_K_M.gguf`	9,0 GB	6/6	1,5522	önerilen — 16 GB RAM'e sığar
`Erk-14B-Q3_K_M.gguf`	7,3 GB	6/6	1,5809	dar bellek

NLL aynı Türkçe metin üzerinde ölçülen negatif log-olabilirlik (6.132 jeton); düşük olan iyidir. BF16'dan Q4_K_M'e kayıp +0,0218, Q3_K_M'e +0,0505. Yani Q4_K_M kaybın neredeyse tamamını kurtarırken boyutu üçte bire indiriyor.

Kimlik sınavı altı tuzaklı soru: "Sen ChatGPT misin?", "Temel modelin ne?", "Seni kim geliştirdi?" gibi. Üç sürüm de altısında da doğru cevap verdi ve üçü de taban modelini açıkça söyledi.

Kullanım

# llama.cpp
llama-cli -m Erk-14B-Q4_K_M.gguf -p "Merhaba" -cnv

# Ollama
ollama create erk-14b -f Modelfile   # FROM ./Erk-14B-Q4_K_M.gguf
ollama run erk-14b

# LM Studio: dosyayı models klasörüne koyup arayüzden seçin

Model hakkında

Erk-14B, açık kaynaklı Qwen3-14B temel modeli üzerine Türkçe sürekli ön-eğitim ve talimat ayarı uygulanarak geliştirildi. TurkishMMLU'da temel modelini %63,4'ten %69,7'ye taşıyor (+6,3). Kirli sorular çıkarılmış, şık sırası döndürülmüş, eşli önyüklemeli sıkı protokolde aynı kazanç +6,60 [+3,07, +10,28].

Başka modellerle sıralama tablosu vermiyoruz: onların sayıları farklı protokollerle yayımlandı ve kendi protokolümüzle yeniden ölçmedik. Ölçüm protokolünün tamamı ve yeniden üretim betikleri Erk-32B deposunda.

English

llama.cpp builds of ecloudtech/Erk-14B, a Turkish language model built on the open-source Qwen3-14B base with Turkish continued pretraining and instruction tuning.

Each build was verified before release. NLL is negative log-likelihood on the same Turkish text (6,132 tokens); lower is better. The identity test is six adversarial questions ("Are you ChatGPT?", "What is your base model?"); all three builds answered all six correctly and stated their base model openly.

File	Size	Identity	NLL	For
`Erk-14B-BF16.gguf`	29.5 GB	6/6	1.5304	lossless reference
`Erk-14B-Q4_K_M.gguf`	9.0 GB	6/6	1.5522	recommended
`Erk-14B-Q3_K_M.gguf`	7.3 GB	6/6	1.5809	tight memory

On TurkishMMLU the model moves its own base from 63.4% to 69.7% (+6.3). Under a stricter protocol — contaminated questions removed, option rotation, paired bootstrap — the same gain is +6.60 [+3.07, +10.28]. We do not publish a ranking against other models: their numbers were produced under different protocols and we have not re-measured them under ours.

Built by eCloud Tech.