Downloads · 30 days
21
5% of all-time downloads
bluehawana/Qwen3.8-27B-Q8-GGUF
Qwen3.8-27B-Q8-GGUF is a machine learning model from bluehawana. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
Single-file Q80 (8-bit, near-lossless) GGUF of Qwen/Qwen3.8-27B — 28.9 GB. Quantized by AtomicChat; re-hosted here alongside our Apple-Silicon serving research for one-command use.
Downloads · 30 days
21
5% of all-time downloads
All-time downloads
409
Public
Repo size
28.9 GB
Likes
0
Public
Click a slice to open those files.
.gguf28.9 GB · 100%
From the Hugging Face model README
Single-file Q8_0 (8-bit, near-lossless) GGUF of Qwen/Qwen3.8-27B — 28.9 GB. Quantized by AtomicChat; re-hosted here alongside our Apple-Silicon serving research for one-command use.
# Ollama (from the Ollama registry — easiest)
ollama run bluehawana/qwen3.8-27b-q8
# Ollama (straight from this repo)
ollama run hf.co/bluehawana/Qwen3.8-27B-Q8-GGUF
# llama.cpp / LM Studio / Jan: download Qwen3.8-27B-Q8_0.gguf directly
Needs ≥48 GB unified memory on Apple Silicon (comfortable at 64 GB+).
This model serves 16 concurrent requests with zero errors on an M-series 128 GB Mac — benchmarks across Ollama / oMLX / SGLang-MLX, plus the SGLang patches that make it possible (upstream PR sgl-project/sglang#35137):
→ bluehawana/qwen3.8-27b-apple-silicon-concurrency
Quick concurrent Ollama serving:
OLLAMA_HOST=127.0.0.1:11500 OLLAMA_NUM_PARALLEL=16 OLLAMA_CONTEXT_LENGTH=8192 ollama serve
# OpenAI-compatible endpoint: http://127.0.0.1:11500/v1
Credits: base model © Qwen (Apache-2.0) · Q8_0 quant by AtomicChat · benchmarks & patches by bluehawana.