GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
wesjos
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
wesjos
Model Tree
Qwen/Qwen3.5-4B
Input Modalities
Output Modalities
Supported Functionality
EN · what is 8.2x9.12/2.3 — right answer, wrong manners · click to enlarge
中文 · python是什么意思 — 问一句被骂一段 · 点击放大
Thumbnails are cropped tight to the chat pane; tap either one for the full-resolution capture.
| Spec | Value |
|---|---|
| Base | Qwen/Qwen3.5-4B (Apache-2.0) |
| Architecture | Qwen3_5ForConditionalGeneration — text + image + video |
| Parameters | 4.66 B (summed from the safetensors headers), bf16 |
| Layers | 32 — 24 Gated DeltaNet (linear attention) + 8 Gated Attention |
| Context | 262,144 tokens native |
| Vocab | 248,320, tied embeddings |
| Shards | 2 × safetensors, 8.68 GiB total |
The rudeness is a fine-tuning effect, not a template trick: the shipped chat template still injects Qwen's default "You are a helpful assistant." persona whenever no system message is present, and the model talks over it anyway.
python
from transformers import AutoProcessor, AutoModelForImageTextToText mid = "wesjos/Qwen3.5-4B-Fuck"processor = AutoProcessor.from_pretrained(mid)model = AutoModelForImageTextToText.from_pretrained(mid, dtype="auto", device_map="auto") messages = [{"role": "user", "content": "what is 8.2x9.12/2.3"}]inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt",).to(model.device) out = model.generate(**inputs, max_new_tokens=512, temperature=0.85, top_p=0.95, top_k=20, presence_penalty=1.5)print(processor.batch_decode(out[:, inputs["input_ids"].shape[-1]:], skip_special_tokens=True)[0])Qwen3.5 needs a current transformers (this checkpoint was exported with 5.10.x). For images, make the user content a list:
python
messages = [{"role": "user", "content": [ {"type": "image", "url": "https://.../cat.jpg"}, {"type": "text", "text": "这张图里有什么？"},]}]→ wesjos/Qwen3.5-4B-Fuck-GGUF — Q4_K_M 2.52 GiB, Q8_0 4.17 GiB, plus an mmproj for vision.
bash
vllm serve wesjos/Qwen3.5-4B-Fuck --port 8000 --max-model-len 32768Qwen3.5 support lands in vLLM / SGLang main; the base card has the full recipe (MTP, tool parsing, thinking mode).
What the demos ran with — warm enough to keep it creative:
| temperature | top_p | top_k | min_p | presence_penalty | repetition_penalty |
|---|---|---|---|---|---|
| 0.85 | 0.95 | 20 | — | 1.5 | off |
Qwen3.5 thinks before answering by default; the screenshots above show the final answer only.
100.7 tok/s throughput readout are as captured.一个被微调成嘴很臭的 Qwen3.5-4B：答案是对的，语气是骂人的。中英文都骂，见上面的 demo 截图。
Apache-2.0, inherited from Qwen3.5-4B. The rude output is the point; shipping it at people who did not ask for it is on you.
from transformers import AutoProcessor, AutoModelForImageTextToText mid = "wesjos/Qwen3.5-4B-Fuck"processor = AutoProcessor.from_pretrained(mid)model = AutoModelForImageTextToText.from_pretrained(mid, dtype="auto", device_map="auto") messages = [{"role": "user", "content": "what is 8.2x9.12/2.3"}]inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt",).to(model.device) out = model.generate(**inputs, max_new_tokens=512, temperature=0.85, top_p=0.95, top_k=20, presence_penalty=1.5)print(processor.batch_decode(out[:, inputs["input_ids"].shape[-1]:], skip_special_tokens=True)[0])messages = [{"role": "user", "content": [ {"type": "image", "url": "https://.../cat.jpg"}, {"type": "text", "text": "这张图里有什么？"},]}]vllm serve wesjos/Qwen3.5-4B-Fuck --port 8000 --max-model-len 32768