Downloads · 30 days
36
4% of all-time downloads
drissea-ai/drissy-qwen3.5-2b-GGUF
drissy-qwen3.5-2b-GGUF is a image-text-to-text model from drissea-ai. Use it for the image-text-to-text 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.
Drissy (pronounced "dris-see") by Drissea
Downloads · 30 days
36
4% of all-time downloads
All-time downloads
856
Public
Repo size
6.5 GB
Likes
2
Public
Click a slice to open those files.
.gguf1.9 GB · 100%
From the Hugging Face model README
Drissy (pronounced "dris-see") by Drissea
GGUF quantized version of Drissy-Qwen3.5-2B for on-device inference with llama.cpp, Ollama, LM Studio, and mobile apps.
| File | Quant | Size | Use Case |
|---|---|---|---|
drissy-qwen3.5-2b-Q4_K_M.gguf | Q4_K_M | ~1.3 GB | Recommended — best balance of size and quality |
llama-cli -m drissy-qwen3.5-2b-Q4_K_M.gguf --jinja --color -ngl 99 -c 8192
llama-server -m drissy-qwen3.5-2b-Q4_K_M.gguf --jinja --port 8080 -ngl 99 -c 8192
from llama_cpp import Llama
llm = Llama(model_path="drissy-qwen3.5-2b-Q4_K_M.gguf", n_ctx=8192, n_gpu_layers=-1)
response = llm.create_chat_completion(messages=[
{"role": "system", "content": "You are Drissy, a private, conversational, and insightful answer engine."},
{"role": "user", "content": "Sources:\n[your chunks here]\n\nUser question: Your question here"}
])
print(response["choices"][0]["message"]["content"])
This model uses Qwen3.5's hybrid Gated DeltaNet architecture. You need the latest version of llama.cpp for full support. Older versions may not recognize the DeltaNet operators.
| Parameter | Value |
|---|---|
| Temperature | 0.7 |
| Top-P | 0.8 |
| Top-K | 20 |
| Context | 8192 (can go up to 262,144) |
| Device | RAM Needed | Speed |
|---|---|---|
| iPhone 12+ (6GB) | ~2 GB | ~5-10 tok/s |
| Mid-range Android (6GB+) | ~2 GB | ~5-8 tok/s |
| MacBook (Apple Silicon) | ~2 GB | ~15-30 tok/s |
| Any PC with GPU | ~2 GB VRAM | ~20-40 tok/s |
This is the quantized deployment version of Drissy, a RAG-grounded answer engine fine-tuned for conversational, source-based answers. See the full model card for benchmark results, architecture details, and training information.
Apache 2.0