You can deploy this model and start earning money today!

jpetrina/
Qwen3.8-27B-MTP-IQ4_XS-pure-GGUF

Name: Qwen3.8-27B-MTP-IQ4_XS-pure-GGUF
Availability: InStock

gguf

unsloth/Qwen3.8-27B-GGUF

Views

No views yet

Qwen3.8-27B MTP Quantized to IQ4_XS for 16 GB VRAM GPUs

This GGUF is basically with identical approach taken by Ununnilium's Qwen3.6-27B-IQ4_XS-pure-GGUF but targetting Multi Token Prediction (MTP) variant of Qwen3.8.

Generated with the bartowski's imatrix like this:

bash

buun-llama-quantize --pure --imatrix Qwen3.8-27B-imatrix.gguf \
    Qwen3.8-27B-BF16-00001-of-00002.gguf qwen3.8-27b-mtp-IQ4_XS-pure.gguf IQ4_XS $(nproc)

There is a non-pure MTP variant generated like this:

bash

buun-llama-quantize --imatrix Qwen3.8-27B-imatrix.gguf \
    --output-tensor-type iq4_xs \
    --tensor-type ".*attn.*=iq4_xs" \
    --tensor-type "nextn=q8_0" \
    Qwen3.8-27B-BF16-00001-of-00002.gguf \
    qwen3.8-27b-mtp-IQ4_XS-Q8nextn.gguf \
    IQ4_XS $$(nproc)

This variant is not strictly speaking --pure since MTP layer weights are quantized down higher to Q8_0 and not IQ4_XS. In other words:

diff

--- a/qwen3.8-27b-mtp-IQ4_XS-pure.gguf-dump
+++ b/qwen3.8-27b-mtp-IQ4_XS-Q8nextn.gguf-dump
@@ -916,7 +916,7 @@
     859:   89128960 | 17408,  5120,     1,     1 | IQ4_XS  | blk.64.ffn_down.weight
     860:   89128960 |  5120, 17408,     1,     1 | IQ4_XS  | blk.64.ffn_gate.weight
     861:   89128960 |  5120, 17408,     1,     1 | IQ4_XS  | blk.64.ffn_up.weight
-    862:   52428800 | 10240,  5120,     1,     1 | IQ4_XS  | blk.64.nextn.eh_proj.weight
+    862:   52428800 | 10240,  5120,     1,     1 | Q8_0    | blk.64.nextn.eh_proj.weight
     863:       5120 |  5120,     1,     1,     1 | F32     | blk.64.nextn.enorm.weight
     864:       5120 |  5120,     1,     1,     1 | F32     | blk.64.nextn.hnorm.weight
     865:       5120 |  5120,     1,     1,     1 | F32     | blk.64.nextn.shared_head_norm.weight

For additional details and recommended parameters see the Unsloth Qwen3.8 Guide.

Qwen3.8-27B

Following the widespread community adoption of the Qwen3.5 and Qwen3.6 series, we are pleased to introduce Qwen3.8, the most capable generation in the Qwen open-model family to date.

Built on the architectural foundation of Qwen3.5, Qwen3.8 delivers substantial gains across coding, professional work, research, and long-horizon agentic tasks. Qwen3.8-27B brings these advances to a compact, deployment-friendly dense model: a native vision-language model that understands images and videos, with flexible thinking control, designed to carry complex, multi-step tasks through to completion with greater reliability.

Qwen3.8 Highlights

Qwen3.8-27B features the following enhancements:

Core Capabilities: Comprehensive improvements across coding, professional work, research, and long-horizon agentic tasks.
Agent Execution: Stronger autonomous planning and better handling of environment feedback, leading to more reliable end-to-end task completion.
Downstream Compatibility: Broader support for popular harnesses and development tools, making it easier to integrate into your existing stack.
Flexible Thinking Control: Thinking mode is on by default and can be disabled per request; reasoning depth can be tuned with reasoning_effort, and reasoning context from historical messages is retained via preserve_thinking.
Vision-Language Understanding: Native support for image and video understanding, from STEM diagrams and documents to hour-scale videos.

Model Overview

Type: Causal Language Model with Vision Encoder
Training Stage: Pre-training & Post-training
Language Model
- Number of Parameters: 27B
- Hidden Dimension: 5120
- Token Embedding: 248,320 (Padded)
- Number of Layers: 64
- Hidden Layout: 16 × (3 × (Gated DeltaNet → FFN) → 1 × (Gated Attention → FFN))
- Gated DeltaNet:
  - Number of Linear Attention Heads: 48 for V and 16 for QK
  - Head Dimension: 128
- Gated Attention:
  - Number of Attention Heads: 24 for Q and 4 for KV
  - Head Dimension: 256
  - Rotary Position Embedding Dimension: 64
- Feed Forward Network:
  - Intermediate Dimension: 17,408
- LM Output: 248,320 (Padded)
- MTP (Multi-Token Prediction): trained with multiple steps
Context Length: 262,144 natively and extensible up to 1,000,000 tokens.

Best Practices

To achieve optimal performance, we recommend the following settings:

Sampling Parameters: We suggest using the following sets of sampling parameters:
- Thinking Mode: temperature=1.0, top_p=0.95, top_k=20, min_p=0.0, presence_penalty=0.0, repetition_penalty=1.0
- Instruct (or non-thinking) mode: temperature=0.7, top_p=0.80, top_k=20, min_p=0.0, presence_penalty=1.5, repetition_penalty=1.0
For supported frameworks, you can adjust the presence_penalty parameter between 0 and 2 to reduce endless repetition. However, using a higher value may occasionally result in language mixing and a slight decrease in model performance.
Adequate Output Length: To optimize performance on agentic tasks, we recommend allocating sufficient output length to allow the model to generate detailed and comprehensive responses. For frameworks that support separate token limits for internal reasoning and final outputs, we suggest the following configuration within the 1M context length:
- Reasoning Content: Set the maximum output length to 262,144 tokens.
- Final Response: Set the maximum output length to 131,072 tokens.
These settings provide the necessary capacity for complex reasoning while ensuring ample space for high-quality final deliverables.
Processing Ultra-Long Texts: Qwen3.8-27B natively supports context lengths of up to 262,144 tokens. For long-horizon tasks where the total length (including both input and output) exceeds this limit, we recommend using RoPE scaling techniques to handle long texts effectively, e.g., YaRN.
Long Video Understanding: To optimize inference efficiency for plain text and images, the size parameter in the released video_preprocessor_config.json is conservatively configured. It is recommended to set the longest_edge parameter in the video_preprocessor_config file to 469,762,048 (corresponding to 224k video tokens) to enable higher frame-rate sampling for hour-scale videos and thereby achieve superior performance. For example,

{"longest_edge": 469762048, "shortest_edge": 4096}

Citation

If you find our work helpful, feel free to give us a cite.

bibtex

@misc{qwen38,
    title = {{Qwen3.8-Max}: A New Bar for Coding and Cowork},
    url = {https://qwen.ai/blog?id=qwen3.8},
    author = {{Qwen Team}},
    month = {August},
    year = {2026}
}

jpetrina/Qwen3.8-27B-MTP-IQ4_XS-pure-GGUF