GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
mrmuminov
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
mrmuminov
Model Tree
NeuronUz/MustaqiLLM
Input Modalities
Output Modalities
Supported Functionality
| Parameter | Value |
|---|---|
| Parameters | 5.17B |
| Architecture | Qwen3ForCausalLM |
| Original architecture | NeuronLMForCausalLM |
| Layers | 36 |
| Hidden size | 3584 |
| Intermediate size | 9728 |
| Attention | GQA |
| Query heads | 28 |
| KV heads | 4 |
| Head dimension | 128 |
| QK normalization | Yes |
| Position encoding | RoPE |
| RoPE θ | 500000 |
| Context length | 4096 tokens |
| Vocabulary | 48,000 BPE |
| Embeddings | Untied |
| Original weights | BF16, with embeddings and lm_head stored as FP32 |
| Languages | Uzbek Latin, Uzbek Cyrillic, English, Russian |
The original MustaqiLLM architecture is functionally equivalent to Qwen3 but stores several projections in fused tensors.
The conversion performs the following mappings:
text
NeuronLM Qwen3------------------------------------------------------qkv_proj.weight -> q_proj.weightk_proj.weightv_proj.weightout_proj.weight -> o_proj.weightq_norm.weight -> q_norm.weightk_norm.weight -> k_norm.weight
The original fused QKV projection has:
text
Q: 28 × 128 = 3584K: 4 × 128 = 512V: 4 × 128 = 512Total QKV output size = 4608
Therefore:
text
qkv_proj├── q_proj: first 3584 rows├── k_proj: next 512 rows└── v_proj: final 512 rows
text
NeuronLM Qwen3------------------------------------------------------gate_up_proj.weight -> gate_proj.weightup_proj.weightdown_proj.weight -> down_proj.weight
gate_up_proj is split equally into the gate and up projections.
All other compatible tensors are copied directly.
This conversion changes the checkpoint layout, not the learned model behavior.
No additional:
was performed as part of the Qwen3 conversion.
Minor numerical differences may occur depending on inference backend, dtype, GPU architecture, batching, and attention implementation.
Install:
bash
pip install -U transformers accelerate torch
Then:
python
import torchfrom transformers import AutoModelForCausalLM, AutoTokenizermodel_id = "mrmuminov/MustaqiLLM-Qwen3"tokenizer = AutoTokenizer.from_pretrained(model_id)model = AutoModelForCausalLM.from_pretrained(model_id,dtype=torch.float16,device_map="auto",).eval()messages = [{"role": "user","content": "O'zbekistonning poytaxti qaysi shahar?"}]inputs = tokenizer.apply_chat_template(messages,add_generation_prompt=True,return_tensors="pt",return_dict=True,).to(model.device)with torch.no_grad():output = model.generate(**inputs,max_new_tokens=256,do_sample=False,)generated = output[0][inputs["input_ids"].shape[1]:]print(tokenizer.decode(generated,skip_special_tokens=True,))
Unlike the original custom architecture, this checkpoint does not require:
python
trust_remote_code=True
One of the main purposes of this conversion is native vLLM compatibility.
Install vLLM:
bash
pip install -U vllm
Start an OpenAI-compatible API server:
bash
vllm serve mrmuminov/MustaqiLLM-Qwen3 \--dtype float16 \--max-model-len 4096
For GPUs with native BF16 support, BF16 may also be used.
For older GPUs such as NVIDIA Turing cards, including the Quadro RTX 8000, use:
bash
--dtype float16
bash
curl http://localhost:8000/v1/chat/completions \-H "Content-Type: application/json" \-d '{"model": "mrmuminov/MustaqiLLM-Qwen3","messages": [{"role": "user","content": "O'\''zbekiston haqida qisqacha gapirib ber."}],"temperature": 0.7,"top_p": 0.9,"max_tokens": 512}'
The native Qwen3 layout should also allow the model to be loaded by Qwen3-compatible SGLang versions.
Example:
bash
python -m sglang.launch_server \--model-path mrmuminov/MustaqiLLM-Qwen3 \--host 0.0.0.0 \--port 30000 \--dtype float16
MustaqiLLM uses a ChatML-style conversation format:
text
<|im_start|>system{system}<|im_end|><|im_start|>user{user}<|im_end|><|im_start|>assistant{assistant}<|im_end|>
Using tokenizer.apply_chat_template() is recommended instead of manually constructing this format.
A system message is optional.
For general conversation, the original model authors recommend omitting a generic system prompt when it is unnecessary.
For open-ended chat, the original MustaqiLLM evaluation recommends approximately:
python
output = model.generate(**inputs,max_new_tokens=512,do_sample=True,temperature=0.7,top_p=0.9,repetition_penalty=1.05,use_cache=True,)
Recommended range:
| Setting | Recommended value |
|---|---|
temperature | 0.7 |
top_p | 0.9 |
repetition_penalty | 1.05–1.10 |
do_sample | True for chat |
For classification, extraction, or very short deterministic answers, greedy decoding can be used.
The original model authors found repetition control particularly important for longer generations.
MustaqiLLM primarily supports:
The model was designed primarily around Uzbek language capabilities.
The following results come from the original MustaqiLLM model card. They have not been independently re-evaluated specifically for this Qwen3-layout conversion.
Because this checkpoint is intended as a layout-preserving conversion of the same model weights, these numbers are included for reference only.
| Benchmark | Samples | Score |
|---|---|---|
| uzlib | 1,861 | 0.2875 |
| TUMLU-Uzbek | 700 | 0.3286 |
| MMLU-Uz | 14,042 | 0.2584 |
| Uzbek news classification | 96,970 | 0.6531 |
| Sentiment classification | 10,000 | 0.9259 |
| Benchmark | Samples | Score |
|---|---|---|
| MMLU | 14,042 | 0.2619 |
| Direction | BLEU | COMET |
|---|---|---|
| English → Uzbek | 5.17 | 0.7397 |
| Uzbek → English | 1.83 | 0.5376 |
See the original model card for the complete evaluation methodology and per-category results:
NeuronUz/MustaqiLLM — Evaluation
The limitations of the original MustaqiLLM model remain applicable to this converted checkpoint.
MustaqiLLM is not primarily a knowledge model.
The original evaluation found performance close to random baseline on several multiple-choice knowledge benchmarks.
It should therefore not be relied upon for:
Using retrieval or external verification is recommended when factual accuracy matters.
Uzbek → English translation is relatively weak according to the original evaluation.
English → Uzbek performs better, but the model should not be considered a replacement for a dedicated machine translation system.
The original model card notes artifacts in Uzbek Cyrillic because some training data was machine-transliterated.
Latin-script Uzbek is generally the safer choice when orthographic accuracy matters.
Long generations may enter repetition loops.
For chat, using:
text
repetition_penalty = 1.05–1.10
is recommended.
The original training data may cause the model to identify itself using an older model or project name.
| Feature | Original | This checkpoint |
|---|---|---|
| Model class | NeuronLMForCausalLM | Qwen3ForCausalLM |
| Custom modeling code | Required | Not required |
trust_remote_code | Required | Not required |
| Fused QKV | Yes | Split |
| Fused Gate/Up | Yes |
This repository contains the converted native Qwen3 checkpoint.
Quantized variants can be produced separately, for example:
Quantized checkpoints should be published separately so that this repository remains the reference Qwen3-compatible conversion.
Suitable uses include:
The model should not be treated as an authoritative factual source.
This checkpoint is derived from:
NeuronUz/MustaqiLLM
Original model:
https://huggingface.co/NeuronUz/MustaqiLLM
All model capability claims and benchmark results above originate from the original MustaqiLLM project unless explicitly stated otherwise.
This repository primarily provides a Qwen3-compatible checkpoint conversion for broader inference-engine compatibility.
The original MustaqiLLM model is released under the Apache License 2.0.
This converted checkpoint follows the same license.
Please review the original repository and license before redistribution or commercial use:
https://huggingface.co/NeuronUz/MustaqiLLM
Thanks to NeuronUz / NeuronAI.uz for developing and releasing MustaqiLLM and making an Uzbek-focused language model publicly available.
The Qwen3 conversion preserves the original model while making the checkpoint easier to deploy using standard inference infrastructure.
use_cacheTrue |
| context | up to 4096 tokens |
| Split |
| Transformers | ✅ | ✅ |
| vLLM native loading | ❌ | ✅ |
| SGLang native loading | ❌ | ✅ |
| Learned weights | MustaqiLLM | MustaqiLLM |
| Additional training | — | None |