Modelpublic

jc-builds/Qwen3.5-0.8B-Q4_K_M-GGUF

sourceHugging Faceapache-2.0updated 13d agoView on Hugging Face

3likes369downloads

Model Card

Qwen3.5-0.8B — GGUF (iPhone-optimized)

A Q4KM GGUF of `Qwen/Qwen3.5-0.8B` for on-device inference on any iPhone, iPad, or Apple Silicon Mac via llama.cpp or apps that wrap it (e.g. Haplo).

Hosted by jc-builds for the Haplo ecosystem. Quantization by Unsloth. Original weights © Alibaba Cloud, redistributed under the Apache 2.0 License.

TL;DR

The smallest model in Alibaba's Qwen3.5 family, using the same hybrid Gated DeltaNet + attention architecture as its larger siblings. At about half a gigabyte it runs on every supported device, and it is a large step up from the previous generation of sub-1B models. It runs in non-thinking mode by default.

Available quantizations

File	Size	Recommended use
`Qwen3.5-0.8B-Q4_K_M.gguf`	0.53 GB	Default — works on every device

Details


Parameters	0.8B
Architecture	`qwen35` (hybrid Gated DeltaNet + attention)
Quantization	Q4KM
Chat format	ChatML with an optional `<think>` block (non-thinking by default)
Minimum device	Any iPhone that runs Haplo

How to use

Haplo (iPhone / iPad / Mac)

The model appears automatically in Haplo's model browser. Download URL:

https://huggingface.co/jc-builds/Qwen3.5-0.8B-Q4_K_M-GGUF/resolve/main/Qwen3.5-0.8B-Q4_K_M.gguf

llama.cpp

bash

llama-cli -hf jc-builds/Qwen3.5-0.8B-Q4_K_M-GGUF:Q4_K_M

License

Apache 2.0. Qwen3.5 by Alibaba Cloud — see the upstream license.