Instructions to use batiai/Qwen3-Embedding-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 batiai/Qwen3-Embedding-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 batiai/Qwen3-Embedding-4B-GGUF:Q6_K
# Run inference directly in the terminal:
llama cli -hf batiai/Qwen3-Embedding-4B-GGUF:Q6_K

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf batiai/Qwen3-Embedding-4B-GGUF:Q6_K
# Run inference directly in the terminal:
llama cli -hf batiai/Qwen3-Embedding-4B-GGUF:Q6_K

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 batiai/Qwen3-Embedding-4B-GGUF:Q6_K
# Run inference directly in the terminal:
./llama-cli -hf batiai/Qwen3-Embedding-4B-GGUF:Q6_K

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 batiai/Qwen3-Embedding-4B-GGUF:Q6_K
# Run inference directly in the terminal:
./build/bin/llama-cli -hf batiai/Qwen3-Embedding-4B-GGUF:Q6_K

Use Docker

docker model run hf.co/batiai/Qwen3-Embedding-4B-GGUF:Q6_K

LM Studio
Jan
Ollama
How to use batiai/Qwen3-Embedding-4B-GGUF with Ollama:
```
ollama run hf.co/batiai/Qwen3-Embedding-4B-GGUF:Q6_K
```
Unsloth Desktop

How to use batiai/Qwen3-Embedding-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 batiai/Qwen3-Embedding-4B-GGUF:Q6_K

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": "batiai/Qwen3-Embedding-4B-GGUF:Q6_K"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use batiai/Qwen3-Embedding-4B-GGUF with Docker Model Runner:
```
docker model run hf.co/batiai/Qwen3-Embedding-4B-GGUF:Q6_K
```

Lemonade

How to use batiai/Qwen3-Embedding-4B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull batiai/Qwen3-Embedding-4B-GGUF:Q6_K

Run and chat with the model

lemonade run user.Qwen3-Embedding-4B-GGUF-Q6_K

List all available models

lemonade list

Hermes Agent

How to use batiai/Qwen3-Embedding-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 batiai/Qwen3-Embedding-4B-GGUF:Q6_K

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 batiai/Qwen3-Embedding-4B-GGUF:Q6_K

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use batiai/Qwen3-Embedding-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 batiai/Qwen3-Embedding-4B-GGUF:Q6_K

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 "batiai/Qwen3-Embedding-4B-GGUF:Q6_K" \
  --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"

Qwen3-Embedding-4B GGUF — Quantized by BatiAI

GGUF quantizations of Qwen/Qwen3-Embedding-4B — the mid-tier sweet spot for 16 GB+ Macs. Higher retrieval quality than 0.6B; leaner RAM than 8B. Part of BatiAI's on-device RAG stack for BatiFlow.

Quick Start

Ollama

ollama pull batiai/qwen3-embedding:4b        # Q6_K (3.3 GB, recommended default)
ollama pull batiai/qwen3-embedding:4b-q8     # Q8_0 (4.3 GB, max quality)

curl http://localhost:11434/api/embeddings -d '{
  "model": "batiai/qwen3-embedding:4b",
  "prompt": "semantic search query"
}'

llama.cpp

./llama-server \
  -m Qwen3-Embedding-4B-Q8_0.gguf \
  --embeddings --pooling last -c 32768 \
  --host 127.0.0.1 --port 8080

Available Quantizations

File	Quant	Size	When to use
`Qwen3-Embedding-4B-Q6_K.gguf`	Q6_K	~3.3 GB	recommended default — Q8↔Q6 drift cos 0.998, indistinguishable on retrieval
`Qwen3-Embedding-4B-Q8_0.gguf`	Q8_0	~4.3 GB	maximum quality, ~25 % bigger disk

Why Q6 over Q8 as default? Measured drift 0.998 = cosine distance well below any retrieval noise floor. Saving ~1 GB matters on 16 GB Macs where every GB of free RAM helps. If you want maximum conservatism, pull :4b-q8.

Why no IQ3 / IQ4 for embedding? Unlike chat LLMs, embedding quality cascades into cosine-similarity drift at low bit-widths. Q6_K / Q8_0 are the safe range.

Quality Verification (measured)

Four-stage harness run on both quants. Full testset + script reproducible via scripts/bench-embedding-quality.sh.

Stage	Test	Q8_0	Q6_K
A. Same-lang semantics	30 (EN+KO) triples, directional correctness	30/30 (100 %)	30/30 (100 %)
	average margin	0.289	0.289
B. Cross-lingual alignment	30 EN↔KO parallel pairs	30/30 (100 %)	30/30 (100 %)
	parallel cos avg	0.762	0.761
	unrelated cos avg	0.220	0.221
	separation Δ	0.542	0.540
C. Real-doc top-1 retrieval	20 EN chunks × 20 EN queries	19/20 (95 %)	19/20 (95 %)
	20 KO chunks × 20 KO queries	20/20 (100 %)	20/20 (100 %)
	combined recall	97.5 %	97.5 %
D. Quant drift	Q8_0 ↔ Q6_K on 20 sample queries	avg cos 0.9984 (min 0.9979, max 0.9990) — PASS

All stages PASS with comfortable margin. Q8_0 and Q6_K produce essentially identical retrieval quality on this testset, with Q6_K saving ~25 % on disk.

Quality tier comparison (across BatiAI text-embedding lineup)

Model	A margin	B separation Δ	C recall (EN / KO)	D drift avg
Qwen3-Embedding-0.6B (Q6)	0.281	0.521	95 % / 100 %	0.9967
Qwen3-Embedding-4B (Q6)	0.289	0.540	95 % / 100 %	0.9984
Qwen3-Embedding-8B (Q6)	0.308	0.569	100 % / 100 %	0.9988

4B sits in the middle — meaningfully better separation than 0.6B, within striking distance of 8B, at roughly half the disk footprint of 8B. Recommended default for 16 GB+ Macs.

Matryoshka — runtime-configurable dimension

Qwen3-Embedding-4B outputs up to 2560 dimensions. BatiFlow RAG stack defaults to 1024 (quality / latency sweet spot per our tests). Truncate at read time — no re-embed needed:

emb = get_embedding(text)     # [2560]
emb_1024 = emb[:1024]         # truncate for storage savings
import numpy as np
emb_1024 = emb_1024 / np.linalg.norm(emb_1024)   # re-normalize

Why text-only?

Qwen3-Embedding-4B is designed specifically for text. For multimodal (image + text), see Qwen3-VL-Embedding-8B on BatiAI.

Task	Use
Document search / Q&A retrieval	this repo (text-only)
Image / screenshot search	`batiai/Qwen3-VL-Embedding-2B-GGUF`

RAG Stack Integration

user query
   ↓ [Qwen3-Embedding 4B]            ← YOU ARE HERE (mid tier)
1024-dim vector
   ↓ vector DB (sqlite-vec / LanceDB)
top-K candidates
   ↓ [Qwen3-Reranker 4B or 8B]
top-3
   ↓ [Qwen3.6-35B-A3B chat LLM]
answer

Recommended Usage — query vs document

# Query side
query = "Instruct: Given a document query, retrieve the most relevant chunk.\n" \
        "Query: " + user_input

# Document side — no instruction prefix
document = chunk_text

BatiFlow handles this automatically. For custom integrations, see Qwen3-Embedding usage.

Why BatiAI?

Quantized directly from Qwen's BF16 safetensors
general.author: BatiAI for provenance
4-stage quality harness (same-lang, cross-lingual, real-doc top-1, quant drift) — numbers published
Paired with Qwen3-Reranker-4B and Qwen3.6-35B-A3B on the same org

Technical Details

Original Model: Qwen/Qwen3-Embedding-4B
Architecture: Qwen3 Causal LM → last-token pooling
Parameters: 4.02 B
Embedding dim: up to 2560 (Matryoshka)
Context: 32 K
License: Apache 2.0
Quantized with: llama.cpp build bafae2765
Quantized by: BatiAI

BatiAI RAG Stack

Role	Model	Repo
Text embedder (entry)	Qwen3-Embedding-0.6B	batiai/Qwen3-Embedding-0.6B-GGUF
Text embedder (mid)	Qwen3-Embedding-4B	this repo
Text embedder (top)	Qwen3-Embedding-8B	batiai/Qwen3-Embedding-8B-GGUF
VL embedder	Qwen3-VL-Embedding-2B / 8B	batiai/Qwen3-VL-Embedding-2B-GGUF
Reranker	Qwen3-Reranker-4B / 8B	batiai/Qwen3-Reranker-4B-GGUF
Chat LLM	Qwen3.6-35B-A3B	batiai/Qwen3.6-35B-A3B-GGUF