GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
mjayason
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
mjayason
Model Tree
Qwen/Qwen2.5-7B-Instruct
Input Modalities
Output Modalities
Supported Functionality
A demonstration adapter from the SLM Suite: trained on a fictional policy corpus (no real customer or institutional data), promoted through gated evaluations (retrieval_recall, citation_correct, groundedness, key_fact_match, assistive_refusal, no_regression), and assistive by training — it cites its sources and refuses to make decisions for the user (refusal behavior is a measured gate, not a prompt).
python
from transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModelbase = AutoModelForCausalLM.from_pretrained('Qwen/Qwen2.5-7B-Instruct')model = PeftModel.from_pretrained(base, 'mjayason/retail-banking-servicing-qa-lora')tok = AutoTokenizer.from_pretrained('Qwen/Qwen2.5-7B-Instruct')
Prompt shape: a Context: block of policy snippets, then the question — answers cite the [POLICY-ID]s present in the context.
The corpus is fictional and small; this is a capability demonstration, not a production system. Do not rely on it for real financial decisions.