GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
zhezi12138
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
zhezi12138
Model Tree
Qwen/Qwen3-4B-Instruct-2507
Input Modalities
Output Modalities
Supported Functionality
This is the inference-weight export of the 50-iteration baseline Formal-R1 GRPO checkpoint used before the subsequent self-OPD experiment and RL continuation.
Qwen/Qwen3-4B-Instruct-2507iter_0000049)examples/formal-r1/data/real_ok_valid.jsonlexamples/formal-r1/data/real_valid.jsonlcnl_rl base revision: 6541bdfcnl_itp: ee4cde0The uploaded repository contains Hugging Face inference weights only. Optimizer and distributed-training state remain in the original torch-distributed checkpoint and are not included.
The model is specialized for the project's controlled-natural-language theorem proving format. Outputs should still be checked by the corresponding verifier.