GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
violetxi
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
violetxi
Model Tree
Qwen/Qwen3.5-9B
Input Modalities
Output Modalities
Supported Functionality
Use a recent Qwen3.5-compatible Transformers/vLLM installation. The frozen evaluations used vLLM 0.21.0; publication metadata is checked with Transformers 5.12.1.
python
from transformers import AutoModelForImageTextToText, AutoTokenizermodel_id = "violetxi/qwen35-9b-wmrl-v4-scale70-3m"tokenizer = AutoTokenizer.from_pretrained(model_id)model = AutoModelForImageTextToText.from_pretrained(model_id, dtype="bfloat16", device_map="auto", use_safetensors=True)
bash
vllm serve violetxi/qwen35-9b-wmrl-v4-scale70-3m --dtype bfloat16 --max-model-len 65536
Agentic 5-turn: 8.04% mean rubric score.
Agentic 20-turn: 7.91% mean rubric score.
Each uses 250 tasks, four samples/task, thinking, the frozen DMS tools/documents
and GLM judge. Closed-book results are in evaluation_summary.json and use
the original probe sampling and scoring. A single training seed was evaluated.
Base revision: c202236235762e1c871ad0ccb60c8ee5ba337b9a. Model parameter/index validation,
file sizes and SHA-256 hashes are in publication_manifest.json.