GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
GParlatto
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
GParlatto
Model Tree
unsloth/Qwen3.5-2B
Input Modalities
Output Modalities
Supported Functionality
Requiere transformers >= 5.2 (soporte de model_type: qwen3_5). El
contenedor por defecto de Inference Endpoints no es compatible: usa
huggingface-inference-toolkit, clavado a transformers==4.51.3.
Ver DEPLOY_VLLM.md para la configuracion de despliegue con vLLM y los riesgos conocidos.
python
from transformers import AutoProcessor, AutoModelForImageTextToTextmodel = AutoModelForImageTextToText.from_pretrained("GParlatto/spoky-qwen-merged-v2",dtype="auto",device_map="auto",)processor = AutoProcessor.from_pretrained("GParlatto/spoky-qwen-merged-v2")messages = [{"role": "user", "content": "Hola, quien eres?"}]inputs = processor.apply_chat_template(messages, tokenize=True, add_generation_prompt=True,return_dict=True, return_tensors="pt",).to(model.device)out = model.generate(**inputs, max_new_tokens=200)print(processor.batch_decode(out[:, inputs["input_ids"].shape[-1]:],skip_special_tokens=True)[0])
La plantilla de chat cierra cada turno con <|im_end|> (248046). El
generation_config.json acepta tanto ese como <|endoftext|> (248044).