GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
willhx
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
willhx
Model Tree
Qwen/Qwen3-8B-Base
Input Modalities
Output Modalities
Supported Functionality
| Stage | Description |
|---|---|
| Base | Qwen/Qwen3-8B-Base |
| SFT | Math + Search cold-start SFT, then Tau (tool-agent) SFT — student init Qwen3-8B-Base-Math-SeaSFT-Search-TauSFT-Tau |
| EOPD | On-policy distillation from a Search specialist teacher (Qwen3-8B-Base-Math-SeaSFT-Search) |
EOPD combines two KL directions to get the best of mode-seeking and mode-covering distillation:
student_logp − teacher_logp is subtracted from the advantage, sharpening the student
toward the teacher's modes.KL(p̃_teacher_topk ‖ p̃_student_topk) is added, gated by the teacher's per-token entropy
(1[H_teacher > τ]) and computed over the teacher's renormalized top-k (k=16) distribution.The teacher is served with SGLang and returns both the scalar sampled-token logprob (reverse
KL) and the per-token top-k distribution + entropy (forward KL). Training was done with the
slime RL framework on Megatron-LM; this repo is the
Hugging Face safetensors export of the final checkpoint.
Identical to Qwen3-8B-Base: 36 layers, hidden size 4096, 32 attention heads, 8 KV heads (GQA), head dim 128, QK-LayerNorm, RMSNorm, SwiGLU, RoPE (θ=1e6), 151,936 vocab, 32,768 context.
python
from transformers import AutoModelForCausalLM, AutoTokenizermodel_id = "willhx/Qwen3-8B-Base-Math-SeaSFT-Search-EOPD-Tau"tok = AutoTokenizer.from_pretrained(model_id)model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="bfloat16", device_map="auto")prompt = "Who won the Nobel Prize in Physics in 1921?"inputs = tok(prompt, return_tensors="pt").to(model.device)out = model.generate(**inputs, max_new_tokens=256)print(tok.decode(out[0], skip_special_tokens=True))
The model is trained for a search/tool-use agent loop (Search-R1 style retrieval); to reproduce the agentic behavior, drive it with the same tool prompt/format used during training.
Apache-2.0, inherited from Qwen3-8B-Base.