GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
teolm30
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
teolm30
Model Tree
unsloth/Qwen3.5-9B
Input Modalities
Output Modalities
Supported Functionality
unsloth/Qwen3.5-9B (Qwen3.5-9B, 9.4B params, multimodal architecture, text decoder fine-tuned)A balanced mix of ~40k samples:
This is a LoRA adapter. Load it on top of the base model with Unsloth or PEFT:
python
from unsloth import FastLanguageModelmodel, tokenizer = FastLanguageModel.from_pretrained(model_name="unsloth/Qwen3.5-9B",max_seq_length=2048,dtype=None,load_in_4bit=True,)model = FastLanguageModel.get_peft_model.from_pretrained(model, "teolm30/ULT-2-max")FastLanguageModel.for_inference(model)
Or with PEFT directly:
python
from transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModelbase = AutoModelForCausalLM.from_pretrained("unsloth/Qwen3.5-9B", load_in_4bit=True)model = PeftModel.from_pretrained(base, "teolm30/ULT-2-max")tokenizer = AutoTokenizer.from_pretrained("unsloth/Qwen3.5-9B")
Apache-2.0 (inherited from Qwen3.5-9B).