Skip to page content
SAVRN
Search Plan a project

Open-weight model · Image and text to text

Qwopus3.8-27B-Flash-GGUF-1M

by Solstice-AI Solstice-AI/Qwopus3.8-27B-Flash-GGUF-1M

Qwopus3.8-27B-Flash-GGUF-1M is an open-weight model for image and text to text from Solstice-AI, released under Apache License 2.0. Its published files total 152.4 GB. It draws 1.2k downloads a month.

Solstice-AI/Qwopus3.8-27B-Flash-GGUF-1M provides the official, production-grade GGUF suite of Qwopus3.8-27B-Flash with native 1,048,576-token (1M) context support, bundled native BF16 multimodal vision projector (mmproj-BF16.gguf), and companion DSpark…

Parameters—
Context—
Weights152.4 GB
Licenseapache-2.0
AccessOpen weights
Monthly Downloads1.2k

Model Card

By Solstice-AI, published under apache-2.0, revision 00589e66f562.

Solstice-AI/Qwopus3.8-27B-Flash-GGUF-1M provides the official, production-grade GGUF suite of Qwopus3.8-27B-Flash with native 1,048,576-token (1M) context support, bundled native BF16 multimodal vision projector (mmproj-BF16.gguf), and companion DSpark drafter. Evaluated under the standardized benchmark harness

Read Solstice-AI's full model card

Qwopus3.8-27B-Flash-1M (GGUF Suite)

Official Solstice-AI Quantization • Native 1M Context Window • Full Multimodal Vision • DSpark Drafters

Model Overview

Solstice-AI/Qwopus3.8-27B-Flash-GGUF-1M provides the official, production-grade GGUF suite of Qwopus3.8-27B-Flash with native 1,048,576-token (1M) context support, bundled native BF16 multimodal vision projector (mmproj-BF16.gguf), and companion DSpark drafter.

Key Specifications

Attribute Specification
Base Model Jackrong/Qwopus3.8-27B-Flash
Architecture Qwen3.5 / Qwopus Conditional Generation with Multimodal Vision
Context Window 1,048,576 tokens (1M native context)
Multimodal Vision Standalone native BF16 projector (mmproj-BF16.gguf)
Bundled Drafter Companion 27B DSpark speculative drafter in speculative/
Target Engines llama.cpp, Ollama, LM Studio, Unsloth

Quantization Ladder & File Matrix

Quant File Size Memory Fit Recommendation / Target
Qwopus3.8-27B-Flash-UD-Q8_K_XL-1M.gguf 30.15 GB 48GB–64GB+ Near-lossless FP16 reference precision
Qwopus3.8-27B-Flash-UD-Q6_K_XL-1M.gguf 23.95 GB 32GB–48GB Extended 6-bit Unsloth Dynamic v3.0
Qwopus3.8-27B-Flash-MTP-Q6_K.gguf 20.89 GB 32GB VRAM Ideal for 32GB GPUs with long context
Qwopus3.8-27B-Flash-MTP-Q5_K_M.gguf 18.19 GB 24GB–32GB Balanced 5-bit high precision
Qwopus3.8-27B-Flash-MTP-Q5_K_S.gguf 17.67 GB 24GB VRAM Compact 5-bit
Qwopus3.8-27B-Flash-UD-Q4_K_XL-1M.gguf 18.42 GB 24GB VRAM High-accuracy 4-bit Unsloth Dynamic
Qwopus3.8-27B-Flash-UD-IQ4_XS-1M.gguf 16.20 GB 16GB–24GB High throughput / tight VRAM limits
mmproj-BF16.gguf 0.87 GB Vision Native vision multimodal projector
speculative/Qwopus3.8-27B-DSpark-Q8_0.gguf 0.88 GB Drafter Speculative decoding companion

Serving Instructions

llama.cpp with DSpark Speculative Decoding & Vision:

llama-cli \
  --hf-repo Solstice-AI/Qwopus3.8-27B-Flash-GGUF-1M \
  --hf-file Qwopus3.8-27B-Flash-MTP-Q6_K.gguf \
  --mmproj mmproj-BF16.gguf \
  --draft-model speculative/Qwopus3.8-27B-DSpark-Q8_0.gguf \
  -c 1048576 \
  -ngl 99

Benchmark Highlights & Validation

Evaluated under the standardized benchmark harness:

Benchmark Suite Discipline Qwopus3.8-27B-Flash (1M) Claude Opus 4.6 Max GPT-4o
SWE-bench Pro Agentic Software Engineering 61.7% 53.4% 48.9%
LiveCodeBench v6 Algorithmic Problem Solving 90.3% 88.8% 72.8%
QwenSWEBench Complex Architecture Refactoring 79.0% 63.8% 61.2%
OSWorld-Verified Desktop & Operating System Automation 84.3% 72.7% 58.7%
ARC-C (Challenge) Frontier Scientific Reasoning 735 (8-Bit) / 719 (4-Bit) ~710–720 63.8%
Long-Context Needle 256K → 1M Tokens Retrieval 100% (Bit-Exact) Pass Pass

Attribution & Acknowledgments

Identity and Version

Repository
Solstice-AI/Qwopus3.8-27B-Flash-GGUF-1M
Publisher
Solstice-AI
Task
Image and text to text
Modality
Image and text
Library
Not stated by the source
Parameters
Not stated by the source
Languages
en, zh
Revision
00589e66f562f9d1affcff874b106e1f51c7e4b1
First published
2026-09-09
Last updated
2026-10-15

Files and Weights

13 files, 152.4 GB in total. The weights are 9 files totalling 152.4 GB in gguf.

Weights9 files · 152.4 GB
Tokenizer1 file · 59.4 KB
Documentation1 file · 4.4 KB
Other1 file · 51.1 KB
Repository1 file · 2.2 KB
Every file
FileTypeSizeSHA-256
Qwopus3.8-27B-Flash-MTP-Q5_K_M.ggufWeights19.5 GB abc97cf74cae
Qwopus3.8-27B-Flash-MTP-Q5_K_S.ggufWeights19.0 GB ed913242473a
Qwopus3.8-27B-Flash-MTP-Q6_K.ggufWeights22.4 GB e41fb287a366
Qwopus3.8-27B-Flash-UD-IQ4_XS-1M.ggufWeights14.3 GB 40fac4050e94
Qwopus3.8-27B-Flash-UD-Q4_K_XL-1M.ggufWeights17.6 GB 3f227079003a
Qwopus3.8-27B-Flash-UD-Q6_K_XL-1M.ggufWeights25.3 GB 701d8fa9ed21
Qwopus3.8-27B-Flash-UD-Q8_K_XL-1M.ggufWeights31.5 GB af36ecb6b5db
mmproj-BF16.ggufWeights931.1 MB 83ee4f4f205f
speculative/Qwopus3.8-27B-DSpark-Q8_0.ggufWeights2.0 GB 7087a8fd05ae
README.mdDocumentation4.4 KB —
chat_template.jinjaOther51.1 KB —
.gitattributesRepository2.2 KB —
tokenizer_config.jsonTokenizer59.4 KB —

License and Download

License
apache-2.0
Access
Open weights, no gate
Download size
152.4 GB
Download from Solstice-AI

Released by Solstice-AI through its official repository on Hugging Face. Read the license.

Built From

  • Derived from Jackrong/Qwopus3.8-27B-Flash
  • Quantized from Jackrong/Qwopus3.8-27B-Flash

Memory Requirements

PrecisionWeights in memory
As published152.4 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Questions About Qwopus3.8-27B-Flash-GGUF-1M

Can I use Qwopus3.8-27B-Flash-GGUF-1M commercially?

Yes. Qwopus3.8-27B-Flash-GGUF-1M is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.

Similar Models

Non-uniform GGUF quantizations of a 512-expert MoE, produced with GSQ and RCO, with a vision projector for multimodal use. This repository provides GGUF quantizations of Qwen3.8-Flash-Next at four sizes, together with the model's vision projector (mmproj) for multimodal use. In contrast to uniform quantization, which applies a single quantization type to all weight tensors, each model here assigns a separate quantization type to every tensor. The assignment is obtained by a gradient-based search that allocates precision according to per-tensor sensitivity, subject to a total size budget. The resulting files are standard GGUF and run unmodified in llama.cpp, Ollama, and LM Studio. A…

Open weights apache-2.0 gguf
View model

Model · Image and text to text

Qwen3.8-27B-iMatrix-NVFP4-MTP-GGUF

Michał Piszczek

I built this quant because the ready-made FP4 file answered the wrong question. It was fast, but on my short WikiText-2 control it scored 6.4949 PPL. Plain Q40 scored 6.3798. The first higher-quality hybrid went too far the other way: good perplexity, 34.19 tok/s, and no comfortable room for 256K plus vision. This is the build that survived both gates. It is a 17.1 GB, 5.01 BPW mixed-precision GGUF of Qwen/Qwen3.8-27B. It keeps large, tolerant matrices in native NVFP4 and spends more bits on selected attention, Gated DeltaNet, and late FFN tensors. The trained MTP layer remains embedded in the same GGUF. This is not a fine-tune. I built the private calibration workload from 5,472 messages…

Open weights apache-2.0
View model

Qwen3.8-27B uncensored by HauhauCS 0/465 Refusals. This is the Aggressive variant: direct answers, no refusal behavior, and minimal preamble on hard prompts. Every text GGUF preserves Qwen3.8's native NextN head, and this release adds HauhauCS FastMTP: a specific acceleration sidecar qualified across the complete quant lineup at maximum native context. Vision is included through the separate BF16 projector. No changes to datasets or intended capabilities. This release preserves Qwen3.8-27B's text, reasoning, agentic, image, and video capabilities while applying the HauhauCS Aggressive uncensoring profile. Pick Aggressive when you specifically want the model to get to the answer without…

Open weights apache-2.0
View model

Model · Image and text to text

Huihui-Qwen3.8-27B-abliterated-GGUF

Huihui.ai

This is an uncensored version of Qwen/Qwen3.8-27B created with abliteration (see remove-refusals-with-transformers to know more about it). This is a crude, proof-of-concept implementation to remove refusals from an LLM model without using TransformerLens. The newly added Huihui-Qwen3.8-27B-abliterated-Underdog-Saluki series come from ConwayResearch/Underdog-Saluki-27B-1.0 and ukisai/Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF. Only layers 22 to 52 (0-based indexing) have been ablated, while the other layers remain unablated. It may come with a small disclaimer warning. Due to ConwayResearch/Underdog-Saluki-27B-1.0 not providing an imatrix file, we will attempt to perform ablation by mixing the GGUF…

Open weights apache-2.0 transformers
View model

Model · Image and text to text

Qwen3.8-27B-GGUF

GGML Org

Run with https://llama.app - https://huggingface.co/Qwen/Qwen3.8-27B - https://huggingface.co/incoai/Qwen3.8-27B-DFlash2 - add info

Open weights apache-2.0
View model

and it does so in 4bit and 8bit. Regular and MTP (fast) NEO IMATRIX GGUFs provided. (this model is part of the Qwen 3.6 27B Fable Fusion 711 pipelines: 2200+ likes, 3 million + downloads) instruct modes (2 new - Spoon / Einstein, all use ZERO REASONING TOKENS) all switchable on the fly via API, direct and "in chat" (yes - model ctrl at the chat/message level). Model name has "plusIQ" in the name. A 12+12 (12 reasoning and 12 instruct) model with interactive optimization/help system will be releasing shortly too. BF16/16-bit MTP GGUF also avail. (there is also a extra robust "tools" version too - Q6 and Q8.) Extreme intelligence in a small package. Jaw dropping performance. Superior…

Open weights apache-2.0
View model