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

bash

# 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.

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

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