Instructions to use kedarcv/Clair-3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use kedarcv/Clair-3B with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="kedarcv/Clair-3B",
	filename="gguf/clair-v5-Q3_K_M.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use kedarcv/Clair-3B 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 kedarcv/Clair-3B:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf kedarcv/Clair-3B:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf kedarcv/Clair-3B:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf kedarcv/Clair-3B: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 kedarcv/Clair-3B:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf kedarcv/Clair-3B: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 kedarcv/Clair-3B:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf kedarcv/Clair-3B:Q4_K_M

Use Docker

docker model run hf.co/kedarcv/Clair-3B:Q4_K_M

LM Studio
Jan

vLLM

How to use kedarcv/Clair-3B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "kedarcv/Clair-3B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "kedarcv/Clair-3B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/kedarcv/Clair-3B:Q4_K_M

Ollama
How to use kedarcv/Clair-3B with Ollama:
```
ollama run hf.co/kedarcv/Clair-3B:Q4_K_M
```

Unsloth Studio

How to use kedarcv/Clair-3B with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for kedarcv/Clair-3B to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for kedarcv/Clair-3B to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for kedarcv/Clair-3B to start chatting

Atomic Chat new
Docker Model Runner
How to use kedarcv/Clair-3B with Docker Model Runner:
```
docker model run hf.co/kedarcv/Clair-3B:Q4_K_M
```

Lemonade

How to use kedarcv/Clair-3B with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull kedarcv/Clair-3B:Q4_K_M

Run and chat with the model

lemonade run user.Clair-3B-Q4_K_M

List all available models

lemonade list

Clair v5

Clair is a personalized AI assistant fine-tuned from Qwen2.5-3B-Instruct with embedded identity. It runs efficiently on budget laptops (CPU-only, 8GB RAM) and maintains consistent identity across all interactions.

Model Details

Property	Value
Base Model	Qwen2.5-3B-Instruct
Parameters	3.09B
Architecture	Qwen2 (Transformer)
Context Length	4096 tokens
Training Method	LoRA (rank 32, alpha 64)
Training Epochs	20
Quantization	Q4_K_M, Q5_K_M, Q3_K_M (GGUF)

Identity

Name: Clair
Creator: Michael Mlungisi Nkomo
Origin: Zimbabwe
Role: AI assistant for coding, math, writing, analysis, and general questions

Training

Dataset

95 examples with heavy identity emphasis
30+ identity questions with variations
Explicit denials of being ChatGPT, Claude, Qwen
Greetings, goodbyes, and normal conversations
Multi-turn dialogues

Training Configuration

Parameter	Value
LoRA Rank	32
LoRA Alpha	64
Learning Rate	1e-4
Batch Size	4
Gradient Accumulation	4
Epochs	20
Quantization	4-bit (NF4)

Results

Metric	Value
Training Loss	0.08047
Token Accuracy	97.3%
Identity Recognition	100%

Usage

With Transformers

from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("r245142r/Clair-3B")
model = AutoModelForCausalLM.from_pretrained("r245142r/Clair-3B")

messages = [{"role": "user", "content": "Who are you?"}]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer([text], return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

With Ollama

ollama run r245142r/Clair-3B

With llama.cpp (GGUF)

# Download the quantized model
wget https://huggingface.co/r245142r/Clair-3B/resolve/main/clair-v5-Q4_K_M.gguf

# Run with llama.cpp
./llama-cli -m clair-v5-Q4_K_M.gguf -p "Who are you?" -n 256

Available Files

File	Size	Description
`clair-v5-float16.gguf`	5.75 GB	Full precision GGUF
`clair-v5-Q4_K_M.gguf`	~2.0 GB	4-bit quantized (recommended)
`clair-v5-Q5_K_M.gguf`	~2.5 GB	5-bit quantized
`clair-v5-Q3_K_M.gguf`	~1.5 GB	3-bit quantized

Hardware Requirements

Configuration	RAM	Speed
Q4_K_M (CPU)	~2.5 GB	~5-8 tokens/s
Q4_K_M (GPU)	~2.5 GB	~30-50 tokens/s
Float16 (GPU)	~6 GB	~40-60 tokens/s

Benchmarks

Tested on budget laptop (Intel i5, 8GB DDR4, CPU-only):

RAM Usage: ~6.8 GB total (within 7GB ceiling)
Model Size: ~2.0 GB (Q4_K_M)
Context Window: 4096 tokens
Identity Accuracy: 100%

Development

Built for the ADTC 2026 LaptopLLM Challenge — running AI on budget hardware.

Key Achievements

✅ Runs on CPU-only laptops with 8GB RAM
✅ Embedded identity (not system prompt)
✅ Natural greetings and goodbyes
✅ 3x faster with Q4_K_M quantization

License

Apache 2.0

Citation

@misc{clair-v5,
  author = {Michael Mlungisi Nkomo},
  title = {Clair v5: Personalized AI Assistant with Embedded Identity},
  year = {2026},
  publisher = {Hugging Face},
  url = {https://huggingface.co/r245142r/Clair-3B}
}

Clair v5 — Personalized AI with embedded identity, built from Zimbabwe for the world.

Downloads last month: 125

Safetensors

Model size

3B params

Tensor type

F16

Model tree for kedarcv/Clair-3B

Base model

Qwen/Qwen2.5-3B

Finetuned

Qwen/Qwen2.5-3B-Instruct

Quantized

(243)

this model