Instructions to use MikaLabs/Vector-L1-4B-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use MikaLabs/Vector-L1-4B-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="MikaLabs/Vector-L1-4B-GGUF",
	filename="Vector-L1-4B-Q4_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 MikaLabs/Vector-L1-4B-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf MikaLabs/Vector-L1-4B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf MikaLabs/Vector-L1-4B-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf MikaLabs/Vector-L1-4B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf MikaLabs/Vector-L1-4B-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 MikaLabs/Vector-L1-4B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf MikaLabs/Vector-L1-4B-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 MikaLabs/Vector-L1-4B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf MikaLabs/Vector-L1-4B-GGUF:Q4_K_M

Use Docker

docker model run hf.co/MikaLabs/Vector-L1-4B-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use MikaLabs/Vector-L1-4B-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "MikaLabs/Vector-L1-4B-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": "MikaLabs/Vector-L1-4B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/MikaLabs/Vector-L1-4B-GGUF:Q4_K_M

Ollama
How to use MikaLabs/Vector-L1-4B-GGUF with Ollama:
```
ollama run hf.co/MikaLabs/Vector-L1-4B-GGUF:Q4_K_M
```

Unsloth Studio

How to use MikaLabs/Vector-L1-4B-GGUF 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 MikaLabs/Vector-L1-4B-GGUF 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 MikaLabs/Vector-L1-4B-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for MikaLabs/Vector-L1-4B-GGUF to start chatting

How to use MikaLabs/Vector-L1-4B-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf MikaLabs/Vector-L1-4B-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "MikaLabs/Vector-L1-4B-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use MikaLabs/Vector-L1-4B-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf MikaLabs/Vector-L1-4B-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 MikaLabs/Vector-L1-4B-GGUF:Q4_K_M

Run Hermes

hermes

Docker Model Runner
How to use MikaLabs/Vector-L1-4B-GGUF with Docker Model Runner:
```
docker model run hf.co/MikaLabs/Vector-L1-4B-GGUF:Q4_K_M
```

Lemonade

How to use MikaLabs/Vector-L1-4B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull MikaLabs/Vector-L1-4B-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Vector-L1-4B-GGUF-Q4_K_M

List all available models

lemonade list

Vector-L1-4B-GGUF

GGUF quantizations of Vector-L1-4B — for running locally with Ollama, LM Studio, llama.cpp, and other local AI runners.

Vector-L1-4B is an open language model built by MikaLabs to help teachers create classroom materials — differentiated worksheets, lesson plans, quizzes, mark schemes, misconception guides, and tailored explanations across Maths and the Sciences.

The "L1" denotes Light, version 1 — the first and smallest member of a planned Vector model family, designed to run on everyday school hardware so teachers can use it locally and offline.

Available Quantizations

File	Quant	Size	Notes
`Vector-L1-4B-Q4_K_M.gguf`	Q4_K_M	~4.8 GB	Recommended. Best balance of quality and size for most machines.

Q4_K_M offers near-full-quality output while staying small enough to run comfortably on modest hardware (a 16 GB GPU runs it with ease; it also runs on CPU with enough system RAM).

Quick Start

Ollama

Option A — run directly from the Ollama library (recommended):

ollama run mikalabs/Vector-L1-4B-GGUF

Library page: https://ollama.com/mikalabs/Vector-L1-4B-GGUF

This is the easiest way to use Vector. The model comes pre-configured with the correct chat template, stop tokens, recommended settings, and system prompt — nothing else to set up.

Option B — build from the GGUF file yourself. Download Vector-L1-4B-Q4_K_M.gguf from this repository, then create a file named Modelfile next to it:

FROM ./Vector-L1-4B-Q4_K_M.gguf

PARAMETER temperature 0.7
PARAMETER top_p 0.8
PARAMETER stop "<|im_end|>"
PARAMETER stop "<|im_start|>"

TEMPLATE """{{ if .System }}<|im_start|>system
{{ .System }}<|im_end|>
{{ end }}{{ if .Prompt }}<|im_start|>user
{{ .Prompt }}<|im_end|>
{{ end }}<|im_start|>assistant
{{ .Response }}<|im_end|>
"""

SYSTEM "You are Vector, a teaching assistant made by MikaLabs that helps educators create worksheets, lesson plans, quizzes, mark schemes, and explanations. You focus on teaching and education."

Then build and run:

ollama create vector-l1 -f Modelfile
ollama run vector-l1

Note: The Modelfile's explicit template and stop tokens are what ensure clean, single-turn responses. Use Option A or B rather than pulling the raw .gguf without a Modelfile.

LM Studio

Download the .gguf file, place it in your LM Studio models folder (or use the in-app downloader), select it, and chat. Set temperature to 0.7 and top_p to 0.8.

llama.cpp

./llama-cli -m Vector-L1-4B-Q4_K_M.gguf -p "Create a differentiated worksheet on Pythagoras' theorem with three tiers, a mark scheme, and common misconceptions. No multiple choice." --temp 0.7 --top-p 0.8

What It's Good At

Vector-L1-4B punches well above its size as a teaching assistant. It excels at:

Differentiated worksheets with genuinely distinct support / core / extension tiers.
Professional mark schemes that separate method marks (M) from answer marks (A).
Subject-specific misconception guides — the actual errors students make, and how to address them.
Structured lesson plans with objectives, starters, main activities, and plenaries.
A wide range of question formats — short answer, true/false, fill-in-the-blank, calculation, explain-your-reasoning — without defaulting to multiple choice.
Strong instruction-following on complex, multi-part requests.
Accurate level calibration for the age or ability you specify.
Clean, ready-to-use output — the resource you asked for, with no filler.

It identifies itself as Vector, a teaching assistant by MikaLabs.

A Note on Scale

Vector-L1-4B is a compact 4-billion-parameter model designed to run on everyday school hardware. It is built for school and secondary-level teaching, not university or research-level material. On very hard problems it may occasionally make mistakes, so — as with any AI tool — answer keys and factual content should be reviewed by a teacher before use with students.

Recommended Settings

Temperature: 0.7
Top-p: 0.8

License

Apache 2.0. Built on Qwen3-4B-Instruct-2507 by the Qwen team, used under the Apache 2.0 license.

Citation

@misc{vector-l1-4b,
  title  = {Vector-L1-4B: An Open Teaching-Assistant Model},
  author = {MikaLabs},
  year   = {2026},
  url    = {https://huggingface.co/MikaLabs/Vector-L1-4B}
}

Downloads last month: 80

GGUF

Model size

4B params

Architecture

qwen3

Hardware compatibility

4-bit

Model tree for MikaLabs/Vector-L1-4B-GGUF

Base model

Qwen/Qwen3-4B-Instruct-2507

Finetuned

unsloth/Qwen3-4B-Instruct-2507

Finetuned

MikaLabs/Vector-L1-4B

Quantized

(2)

this model