GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
lixiaochuan2020
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
lixiaochuan2020
Model Tree
Qwen/Qwen3-1.7B
Input Modalities
Output Modalities
Supported Functionality
Public: anyone can download it, no access request or login needed.
dtype=torch.float32, or cast to bfloat16 for inference.main: the final checkpoint, step 4,763 of 4,763. Intermediate checkpoints, if added later, go on branches named step-<N>.A Qwen/Qwen3-32B teacher (thinking mode) answered NQ and HotpotQA training questions as a search agent, 4 times per question, using a wiki_search tool over wiki-18 (e5-base-v2, 5 passages per search, at most 10 searches, at most 16,384 tokens per trajectory). Trajectories were kept if they called at least one search and the Qwen3-32B answer judge marked the final answer correct. Gold answers were used only for that check, never as training text.
| Item | Value |
|---|---|
| Questions answered | 128,875 |
| Trajectories kept | 297,720 |
| Supervised (assistant) tokens | 243.78M |
| Tool-result (passage) tokens | 314.24M |
| Total tokens | 684.86M |
The loss covers assistant tokens only.
| Setting | Value |
|---|---|
| Base model | Qwen/Qwen3-1.7B |
| Optimizer | Muon for the decoder layers' weight matrices, AdamW for embeddings and norms |
| Learning rate | 2e-5 |
| Schedule | cosine, 3% warmup |
| Epochs | 1 |
| Steps | 4,763 |
| Packed row length (tokens) | 18,432 (padding-free; each trajectory attends only to itself) |
| Rows per step | 8 (8 GPUs × 1 row) |
| Weight decay | 0 |
| Max gradient norm | 1.0 |
| Precision | bf16 autocast, float32 master weights |
Full causal attention, as in stock Qwen3. model.generate() and stock vLLM/SGLang run it as it was trained.
Qwen3 chat template with one tool, wiki_search, called in Qwen3's <tool_call> format. Each tool result lists the top 5 wiki-18 passages as Doc k(Title: …) text lines. The system prompt and tool schema used in training are SYSTEM_PROMPT and TOOLS in src/kv_midtrain/data_loading/afm_mhqa.py of ashmitkx/kv-midtrain (private repository).
Only measured numbers; "pending" where an eval has not been run.
| Eval | Setting | Metric | Value |
|---|---|---|---|
| Held-out search QA: 1,200 questions (720 HotpotQA dev, 240 NQ test, 240 2WikiMultihopQA dev), 3 samples each | self-generate with live wiki-18 search, temperature 0.7, top-p 0.95 | Judge accuracy (Qwen3-32B, final answer) | pending |
| same | same | Exact match | 0.333 |
| same | same | F1 | 0.425 |
| GSM8K test (1,319 problems) | no tools, 1 sample, temperature 0.6, top-p 0.95, top-k 20 | Judge accuracy (Qwen3-32B, final \boxed{} answer) | 0.660 |
| MATH-500 (500 problems) | same | same | 0.646 |
| Out-of-distribution search: Bamboogle, MuSiQue, PopQA, FEVER (1,025 questions) | self-generate with live wiki-18 search, temperature 0.7, top-p 0.95 | Judge accuracy (Qwen3-32B, final answer) | pending |