Instructions to use Bedirhn77/Mithra-Evolve-4B-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 Bedirhn77/Mithra-Evolve-4B-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 Bedirhn77/Mithra-Evolve-4B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Bedirhn77/Mithra-Evolve-4B-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 Bedirhn77/Mithra-Evolve-4B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Bedirhn77/Mithra-Evolve-4B-GGUF:Q4_K_M

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf Bedirhn77/Mithra-Evolve-4B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama cli -hf Bedirhn77/Mithra-Evolve-4B-GGUF:Q4_K_M

Use Docker

docker model run hf.co/Bedirhn77/Mithra-Evolve-4B-GGUF:Q4_K_M

Llama new
LM Studio
Jan

vLLM

How to use Bedirhn77/Mithra-Evolve-4B-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Bedirhn77/Mithra-Evolve-4B-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": "Bedirhn77/Mithra-Evolve-4B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Bedirhn77/Mithra-Evolve-4B-GGUF:Q4_K_M

Ollama
How to use Bedirhn77/Mithra-Evolve-4B-GGUF with Ollama:
```
ollama run hf.co/Bedirhn77/Mithra-Evolve-4B-GGUF:Q4_K_M
```
Unsloth Desktop

How to use Bedirhn77/Mithra-Evolve-4B-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Bedirhn77/Mithra-Evolve-4B-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": "Bedirhn77/Mithra-Evolve-4B-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use Bedirhn77/Mithra-Evolve-4B-GGUF with Docker Model Runner:
```
docker model run hf.co/Bedirhn77/Mithra-Evolve-4B-GGUF:Q4_K_M
```

Lemonade

How to use Bedirhn77/Mithra-Evolve-4B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Bedirhn77/Mithra-Evolve-4B-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Mithra-Evolve-4B-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use Bedirhn77/Mithra-Evolve-4B-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 Bedirhn77/Mithra-Evolve-4B-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 Bedirhn77/Mithra-Evolve-4B-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use Bedirhn77/Mithra-Evolve-4B-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Bedirhn77/Mithra-Evolve-4B-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 "Bedirhn77/Mithra-Evolve-4B-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"

Mithra-Evolve-4B (GGUF, Q4_K_M)

Experimental v0.1 of Mithra Evolve: a small coding/agent model meant to run on CPU through Ollama with low RAM (file size about 2.8 GB) and to be used from Claude Code-style agent workflows.

Status: early prototype. It is a QLoRA SFT of Qwen/Qwen3.5-4B on a few thousand examples. It is not claimed to match Claude Haiku, Sonnet or Opus. Only the numbers below were measured.

Measured results

Benchmark	Base (`Qwen/Qwen3.5-4B`)	Mithra-Evolve-4B
HumanEval pass@1 (greedy, non-thinking, 512 new tokens, 4-bit load)	80.5% (n=164)	71.3% (n=164)

HumanEval is a weak proxy for agentic work. SWE-bench, Terminal-Bench and LiveCodeBench have not been run yet.

Training

Method: QLoRA SFT (r=32), assistant-only loss, max sequence 4096, 1 epoch, fp16 on Kaggle T4.
Data (resolved trajectories only for agent data; examples longer than 4096 tokens dropped):
nebius/SWE-agent-trajectories: 244 examples
SWE-bench/SWE-smith-trajectories: 0 examples
ise-uiuc/Magicoder-OSS-Instruct-75K: 1500 examples
Decontamination: 13-gram overlap filter against HumanEval and MBPP.
Please check the licenses of the source datasets before any commercial use.

Use with Ollama

ollama run hf.co/Bedirhn77/Mithra-Evolve-4B-GGUF:Q4_K_M

For tool calling (Claude Code), the chat template matters. If tools are not called correctly, copy the TEMPLATE/PARSER lines from ollama show qwen3.5:4b --modelfile into your own Modelfile (FROM ./Mithra-Evolve-4B-Q4_K_M.gguf).

LoRA adapter: https://huggingface.co/Bedirhn77/Mithra-Evolve-4B-LoRA

Limitations

Small model, short training, English-centric data, no safety tuning beyond the base model. Verify generated code.

Downloads last month: 67

GGUF

Model size

4B params

Architecture

qwen35

Hardware compatibility

4-bit

Model tree for Bedirhn77/Mithra-Evolve-4B-GGUF

Base model

Qwen/Qwen3.5-4B-Base

Finetuned

Qwen/Qwen3.5-4B

Quantized

(604)

this model