Downloads · 30 days
398
6% of all-time downloads
localweights/Qwen3.6-35B-A3B-MTP-IMAT-IQ4_XS-Q8nextn-GGUF
Qwen3.6-35B-A3B-MTP-IMAT-IQ4_XS-Q8nextn-GGUF is a text generation model from localweights. Use it when you need the model to write or continue text. The card lists the license as apache-2.0.
35B MoE Qwen3.6-A3B trunk (3B active) + embedded NextN-MTP head, quantized for single-GPU inference.
Downloads · 30 days
398
6% of all-time downloads
All-time downloads
6.4K
Public
Repo size
19.2 GB
Likes
5
Public
Click a slice to open those files.
.gguf19.2 GB · 100%
From the Hugging Face model README
35B MoE Qwen3.6-A3B trunk (3B active) + embedded NextN-MTP head, quantized for single-GPU inference.
kv_only_nextn=true)The MTP head is embedded inside the GGUF — no separate drafter file. Recent llama.cpp builds activate it via --spec-type draft-mtp.
llama-server -m Qwen3.6-35B-A3B-MTP-IMAT-IQ4_XS-Q8nextn.gguf \
-ngl 999 -fa on -c 32768 --parallel 1 \
-ctk q8_0 -ctv q8_0 --kv-unified \
--host 0.0.0.0 --port 8080 --jinja
Requires llama.cpp built from master after ggml-org/llama.cpp#22673:
llama-server -m Qwen3.6-35B-A3B-MTP-IMAT-IQ4_XS-Q8nextn.gguf \
-ngl 999 -fa on -c 32768 --parallel 1 \
-ctk q8_0 -ctv q8_0 --kv-unified \
--spec-type draft-mtp \
--spec-draft-n-max 2 \
--spec-draft-p-min 0.0 \
--host 0.0.0.0 --port 8080 --jinja
For A3B MoE, smaller --spec-draft-n-max wins — accept rate decays faster with depth than dense models. n_max=2 is the sweet spot. For MoE at deeper chains (n>=3), --spec-draft-p-min 0.0 outperforms p_min=0.75 because per-step sampler overhead from top_k+softmax exceeds the accept-rate savings on this arch.
Single-stream decode on a 24 GB consumer GPU (RTX 3090 Ti), CUDA build, full-GPU offload, FA on, N=512 generated tokens, 5-warm-run mean.
| config | decode tok/s |
|---|---|
| no spec (greedy) | ~120 (varies) |
--spec-type draft-mtp --spec-draft-n-max 2 --spec-draft-p-min 0.0 | ~200 (+67%) |
| n_max | tok/s |
|---|---|
| 1 | 189.5 |
| 2 | 198.7 (+24%) |
| 3 | 191.2 |
| 4 | 167.6 |
n_max=2 wins; deeper chains regress (MoE accept rate falls faster with depth).
IQ4_XS calibrated with an imatrix derived from a mixed-domain calibration set.Q8_0 (NextN). Sensitive to precision.Q8_0/Q8_0 recommended.LLM_ARCH_QWEN35MOE_MTP). Merged upstream in #22673.--spec-draft-p-min 0.75 requires recent llama.cpp that honors p_min on the DRAFT_MTP path.kv_only_nextn=true.Apache 2.0 license.