GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
Rumiii
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
Rumiii
Model Tree
Qwen/Qwen3-8B
Input Modalities
Output Modalities
Supported Functionality
enable_thinking=False); retrieved <observation> text is masked from the lossThis is NOT native Qwen tool calling. Run a loop: generate until </wiki_search>, run your own retriever,
append <observation> ... </observation>, and continue until </answer>.
Use the dataset's system prompt (see the dataset card) and enable_thinking=False in the chat template.
python
from transformers import AutoTokenizer, AutoModelForCausalLMtok = AutoTokenizer.from_pretrained("Rumiii/Qwen3-8B-Seeker")model = AutoModelForCausalLM.from_pretrained("Rumiii/Qwen3-8B-Seeker", dtype="auto", device_map="auto")msgs = [{"role": "user", "content": "Who are you?"}]p = tok.apply_chat_template(msgs, tokenize=False, add_generation_prompt=True, enable_thinking=False)out = model.generate(**tok(p, return_tensors="pt").to(model.device), max_new_tokens=200)print(tok.decode(out[0], skip_special_tokens=True))Fine-tuned on a subset of one dataset; it has no built-in search tool and needs an external retriever for the agent loop.