Instructions to use ekovshilovsky/Qwen2.5-32B-TQ8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ekovshilovsky/Qwen2.5-32B-TQ8 with MLX:

# Make sure mlx-lm is installed
# pip install --upgrade mlx-lm

# Generate text with mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("ekovshilovsky/Qwen2.5-32B-TQ8")

prompt = "Write a story about Einstein"
messages = [{"role": "user", "content": prompt}]
prompt = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True
)

text = generate(model, tokenizer, prompt=prompt, verbose=True)

Notebooks
Google Colab
Kaggle
Local Apps Settings
LM Studio

How to use ekovshilovsky/Qwen2.5-32B-TQ8 with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "ekovshilovsky/Qwen2.5-32B-TQ8"

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "ekovshilovsky/Qwen2.5-32B-TQ8"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use ekovshilovsky/Qwen2.5-32B-TQ8 with Hermes Agent:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "ekovshilovsky/Qwen2.5-32B-TQ8"

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 ekovshilovsky/Qwen2.5-32B-TQ8

Run Hermes

hermes

MLX LM

How to use ekovshilovsky/Qwen2.5-32B-TQ8 with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "ekovshilovsky/Qwen2.5-32B-TQ8"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "ekovshilovsky/Qwen2.5-32B-TQ8"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "ekovshilovsky/Qwen2.5-32B-TQ8",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Qwen2.5-32B-TQ8

TurboQuant-compressed version of Qwen/Qwen2.5-32B for near-lossless inference on Apple Silicon.

Compressed with turboquant-mlx-core using the TurboQuant algorithm (Zandieh et al., ICLR 2026).

Quality

Metric	Value
fp16 PPL	1.41
TQ8 PPL	1.42
PPL delta	0.09%
Compression	56% of original size

Quantization Config

Property	Value
Method	TurboQuant 4+4 residual (8 effective bits)
Rotation	Dual Walsh-Hadamard (per-pass seeds)
Codebooks	Per-layer Lloyd-Max fitted
Sensitive layers	First/last 4 at fp16
Block size	Adaptive (largest power-of-2 dividing in_features)

Usage

# Serve via SwiftLM (dequants to BF16 on first load, cached for subsequent runs)
SwiftLM --model ekovshilovsky/Qwen2.5-32B-TQ8 --port 5413

# Dequant to fp16 for use with any MLX/HuggingFace loader
tq-dequant ./Qwen2.5-32B-TQ8 ./Qwen2.5-32B-fp16

Hardware Requirements

Apple Silicon Mac (M1 Pro+ recommended)
64 GB unified memory minimum (34 GB model + KV cache + overhead)
macOS 14+

Original Model

This is a quantized version of Qwen/Qwen2.5-32B by Alibaba Cloud. The original model is released under the Apache 2.0 License. All original model terms and conditions apply.

Quantization

Quantization performed by Eugene Kovshilovsky using turboquant-mlx-core (MIT License).

Downloads last month: 72

Safetensors

Model size

58B params

Tensor type

F32

U32

MLX

Hardware compatibility

4-bit

Model tree for ekovshilovsky/Qwen2.5-32B-TQ8

Base model

Qwen/Qwen2.5-32B

Quantized

(78)

this model