GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
Abdullah452
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
Abdullah452
Model Tree
Qwen/Qwen2.5-0.5B-Instruct
Input Modalities
Output Modalities
Supported Functionality
The draft model for TrialLens' greedy speculative decoding. It was fine-tuned on the same training split and with the same recipe as the target adapter (LoRA r=16, 8.8M parameters, epoch 3 of 4, validation accuracy 0.738), so its proposals match the target's answer format.
On the criterion task it raised draft acceptance from 60 to 83% (stock 0.5B) to 97 to 100%. Speculative decoding never changes the output: every committed token is verified by the 1.5B target.
Educational prototype, not for clinical use. Code: https://github.com/AbdullahRasheed452/TrialLens