GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
nabin2004
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
nabin2004
Model Tree
google/gemma-4-31B-it
Input Modalities
Output Modalities
Supported Functionality
| Artifact | Repo |
|---|---|
| LoRA adapter | nabin2004/AOS-gemma4-31b-manim-sft |
| GGUF (Ollama / llama.cpp) | nabin2004/AOS-gemma4-31b-manim-gguf |
Merged from google/gemma-4-31B-it and the AOS Manim SFT LoRA adapter. Accept the Gemma license and set HF_TOKEN to download.
Fine-tuned on nabin2004/AOS-Trajectories using the AOS Phase 1 SFT trainer (apps/sft).
python
import osimport torchfrom transformers import AutoModelForImageTextToText, AutoTokenizertoken = os.environ["HF_TOKEN"]model_id = "nabin2004/AOS-gemma4-31b-manim-merged"tokenizer = AutoTokenizer.from_pretrained(model_id, token=token)model = AutoModelForImageTextToText.from_pretrained(model_id,torch_dtype=torch.bfloat16,device_map="auto",token=token,)
Serve as a regular model (no --enable-lora needed):
bash
vllm serve nabin2004/AOS-gemma4-31b-manim-merged --max-model-len 16384
Convert to GGUF or pull the pre-built GGUF repo. See apps/sft/export_gguf.py.
bash
cd apps/sftuv run python merge_adapter.py \--adapter-dir ./gemma4-31b-manim-ft \--output-dir ./gemma4-31b-manim-merged \--push-to-hub