GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
ssurface
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
ssurface
Model Tree
Qwen/Qwen3-4B-Instruct-2507
Input Modalities
Output Modalities
Supported Functionality
text
Qwen/Qwen3-4B-Instruct-2507 → SFT LoRA (ssurface/qwen3-4b-cot-compress-l5) → Merged → GRPO with DAPO loss → Merged (this model)Two reward signals were used during GRPO:
Checks if the model's final answer (after ####) matches the ground truth.
Reward is scaled by problem complexity (1–5 steps):
| Complexity | Reward |
|---|---|
| 5 (hardest) | 5.0 |
| 4 | 4.0 |
| 3 | 3.0 |
| 2 | 2.0 |
| 1 (easiest) | 1.0 |
| Wrong answer | 0.0 |
Penalizes <think> blocks that are too short OR too long for the target level.
| Level | Style | Target range (chars) | Too-short penalty | Too-long penalty |
|---|---|---|---|---|
| 1 | Verbose | 120–240 | −0.05/char | −0.01/char |
| 2 | Concise | 50–110 | −0.05/char | −0.01/char |
| 3 | Symbolic | 40–100 | −0.05/char | −0.01/char |
| 4 | Shorthand | 20–60 | −0.05/char | −0.01/char |
| 5 | Extreme | 10–30 | −0.05/char | −0.01/char |
This model targets Level 5 (Extreme): 10-30 chars of reasoning.
| Parameter | Value |
|---|---|
| Loss type | DAPO |
| LoRA r | 16 |
| LoRA alpha | 32 |
| Num generations | 8 |
| Max completion length | 256 |
| Batch size | 16 |
| Gradient accumulation | 2 |
| Learning rate | 1e-5 |
| Epochs | 1 |
| Dataset | GSM8K (train split, level-filtered) |
python
from transformers import AutoModelForCausalLM, AutoTokenizer model = AutoModelForCausalLM.from_pretrained("ssurface/qwen3-4b-grpo-l5")tokenizer = AutoTokenizer.from_pretrained("ssurface/qwen3-4b-grpo-l5") messages = [{"role": "user", "content": "Solve this using Level 5 (Extreme).\nProblem: ..."}]inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt")outputs = model.generate(inputs, max_new_tokens=256)print(tokenizer.decode(outputs[0], skip_special_tokens=False))Qwen/Qwen3-4B-Instruct-2507 → SFT LoRA (ssurface/qwen3-4b-cot-compress-l5) → Merged → GRPO with DAPO loss → Merged (this model)from transformers import AutoModelForCausalLM, AutoTokenizer model = AutoModelForCausalLM.from_pretrained("ssurface/qwen3-4b-grpo-l5")tokenizer = AutoTokenizer.from_pretrained("ssurface/qwen3-4b-grpo-l5") messages = [{"role": "user", "content": "Solve this using Level 5 (Extreme).\nProblem: ..."}]inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt")outputs = model.generate(inputs, max_new_tokens=256)print(tokenizer.decode(outputs[0], skip_special_tokens=False))