GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
L1nkee
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
L1nkee
Model Tree
Qwen/Qwen3-4B-Base
Input Modalities
Output Modalities
Supported Functionality
DPPO binary-TV, thinking, KL 0, LR 1e-6, mini-batch 32, fp32 actor, 540 steps (T20).
RL fine-tune of Qwen/Qwen3-4B-Base on DAPO-Math with verl, from a study of policy-update budgets (ODBA). Weights are bf16 safetensors merged from FSDP checkpoints.
main holds the final step (540). Every saved step is a branch named step-<N>: 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 130, 135, 140, 150, 160, 170, 180, 190, 200, 210, 220, 230, 240, 250, 260, 270, 280, 290, 300, 310, 320, 330, 340, 350, 360, 370, 380, 390, 400, 410, 420, 430, 440, 450, 460, 470, 480, 490, 500, 510, 520, 530, 540.
python
from transformers import AutoModelForCausalLMmodel = AutoModelForCausalLM.from_pretrained("L1nkee/Qwen3-4B-Base-DPPO-BinaryTV-LR1e-6-FP32-T20", revision="step-10")