creekhop/Qwen3.8-27b-EXP-EVE-GGUF overview
Qwen3.8 27b EXP EVE GGUF GGUF quantizations of: win10/Qwen3.8 27b EXP EVE https://huggingface.co/win10/Qwen3.8 27b EXP EVE Source Model Card This is a successf…
Runs locally from ~15.02 GB disk (16 GB VRAM class GPUs with llama.cpp / guIDE).
Repository Files & Downloads
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| Qwen3.8-27b-EXP-EVE-BF16.gguf | GGUF | BF16 | 50.90 GB | Download |
| Qwen3.8-27b-EXP-EVE-IQ4_NL.gguf | GGUF | IQ4_NL | 15.02 GB | Download |
| Qwen3.8-27b-EXP-EVE-Q4_K_M.gguf | GGUF | Q4_K_M | 15.66 GB | Download |
| Qwen3.8-27b-EXP-EVE-Q5_K_M.gguf | GGUF | Q5_K_M | 18.19 GB | Download |
| Qwen3.8-27b-EXP-EVE-Q6_K.gguf | GGUF | Q6_K | 20.89 GB | Download |
| Qwen3.8-27b-EXP-EVE-Q8_0.gguf | GGUF | Q8_0 | 27.05 GB | Download |
Model Details
Model README
---
license: apache-2.0
base_model:
- win10/Qwen3.8-27b-EXP-EVE
base_model_relation: quantized
library_name: llama.cpp
tags:
- gguf
- qwen
- llama.cpp
- quantized
- win10
- Qwen3.8-27b-EXP-EVE
---
Qwen3.8-27b-EXP-EVE-GGUF
GGUF quantizations of:
Source Model Card
This is a successfully merged cross-architecture model, built around the popular Qwen3.8-27B as its primary backbone.
Using a Tensor Gene Evolution merging approach, I incorporated capabilities and behavioral characteristics from meta-models/Muse-Glimmer-30B and google/gemma-4-31B-it into the Qwen backbone.
Compared with the original base model, this merged model appears to preserve the donors' reasoning characteristics more effectively, while demonstrating broader reasoning coverage, greater depth of thought, and more diverse problem-solving behavior.
In my observations, its reasoning ability can be surprisingly strong and, in some cases, may even appear more capable than DeepSeek models. Further systematic evaluation and benchmarking are still needed to quantify these differences.
💖 Open Sponsorship
Support this research and future model development:
Run creekhop/Qwen3.8-27b-EXP-EVE-GGUF with guIDE
Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.
Source: Hugging Face · Compare models