Instructions to use Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP",
	filename="Qwen3.5-35B-A3B-IQ3_S-MTP.gguf",
)

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

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S
# Run inference directly in the terminal:
llama-cli -hf Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S
# Run inference directly in the terminal:
llama-cli -hf Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S

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 Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S
# Run inference directly in the terminal:
./llama-cli -hf Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S

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 Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S

Use Docker

docker model run hf.co/Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S

LM Studio
Jan

vLLM

How to use Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S

Ollama
How to use Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP with Ollama:
```
ollama run hf.co/Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S
```

Unsloth Studio new

How to use Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP 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 Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP 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 Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP to start chatting

Pi new

How to use Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S

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": "Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S

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 Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S

Run Hermes

hermes

Docker Model Runner
How to use Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP with Docker Model Runner:
```
docker model run hf.co/Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S
```

Lemonade

How to use Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP:IQ3_S

Run and chat with the model

lemonade run user.Qwen3.5-35B-A3B-IQ3_S-MTP-IQ3_S

List all available models

lemonade list

Qwen3.5-35B-A3B — IQ3_S with MTP Tensors

One of the first GGUF quantizations of Qwen3.5-35B-A3B that preserves the Multi-Token Prediction (MTP) head. Most public quantizations strip MTP tensors during conversion. This GGUF retains them, enabling speculative decoding via MTP.

Key specs

Metric	Value
Base model	Qwen/Qwen3.5-35B-A3B
File size	11 GB
Quantization	IQ3_S custom-mix (base) + BF16 MTP tensors
Total layers	41 (40 main + 1 MTP)
Total tensors	753
MTP acceptance	84.8% (warm cache), 61-63% (cold)

MTP Results

Tested with ik_llama.cpp (patched for Qwen3.5 MoE MTP support):

Config	Speed	Acceptance	Notes
Without `-mtp`	94 tok/s	—	Normal inference
`-mtp --draft-max 1`	47.6 tok/s	84.8% warm / 61% cold	Works but slower
`-mtp` (multi-step)	12.7 tok/s	12%	Acceptance collapse (known MTP issue)

Why MTP is slower here

The MTP layer (layer 40) is itself a full MoE layer with 256 experts. On consumer hardware with CPU expert offloading (-ncmoe 4), each MTP forward costs nearly as much as a main model forward. The bottleneck is MoE expert transfer over PCIe, not the number of forward passes.

This is an informative negative result: MTP speculative decoding is not cost-effective on MoE models where the MTP head is also MoE. It would work on models with a dense MTP head.

MTP patches for ik_llama.cpp

This GGUF requires patched ik_llama.cpp with MTP support for Qwen3.5 MoE. The 5 patches (fixing 8 bugs in tensor mapping, compute buffers, and MoE handling) are available at: chimere/patches/ik-llama-mtp

How it was built

# Inject MTP tensors from original BF16 shards into IQ3_S base
python build_mtp_gguf_v3.py
# Source: IQ3_S custom-mix base + shards 13-14 from Qwen/Qwen3.5-35B-A3B

Build tool: ramp-quant/tools/build_mtp_gguf_v3.py

chimere — Rust inference runtime + MTP patches
Qwen3.5-35B-A3B-RAMP-v2-15G — Optimized quantization without MTP

Author

Kevin Remondiere — Independent ML researcher, Oloron-Sainte-Marie, France

License

Apache 2.0. Base model follows Qwen's license.

Downloads last month: 45

GGUF

Model size

37B params

Architecture

qwen35moe

Hardware compatibility

3-bit

Model tree for Kevletesteur/Qwen3.5-35B-A3B-IQ3_S-MTP

Base model

Qwen/Qwen3.5-35B-A3B-Base

Finetuned

Qwen/Qwen3.5-35B-A3B

Quantized

(258)

this model

Kevletesteur
/

Qwen3.5-35B-A3B-IQ3_S-MTP