ertghiu256/Qwen3-4b-tcomanr-merge-v2.3

Ties merged COde MAth aNd Reasoning model

This is a merge of pre-trained language models created using mergekit.

Merge Details

This model is a revision of the ertghiu256/Qwen3-4b-tcomanr-merge-v2.2

This model aims to combine the reasoning, code, and math capabilities of Qwen3 4b 2507 reasoning by merging it with some other Qwen3 finetunes. This model reasoning is very long.

How to run

You can run this model by using multiple interface choices

Transformers

As the qwen team suggested to use

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "ertghiu256/Qwen3-4b-tcomanr-merge-v2.3"

# load the tokenizer and the model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)

# prepare the model input
prompt = "Give me a short introduction to large language model."
messages = [
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

# conduct text completion
generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=32768
)
output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist() 

# parsing thinking content
try:
    # rindex finding 151668 (</think>)
    index = len(output_ids) - output_ids[::-1].index(151668)
except ValueError:
    index = 0

thinking_content = tokenizer.decode(output_ids[:index], skip_special_tokens=True).strip("\n")
content = tokenizer.decode(output_ids[index:], skip_special_tokens=True).strip("\n")

print("thinking content:", thinking_content) # no opening <think> tag
print("content:", content)

Vllm

Run this command

vllm serve ertghiu256/Qwen3-4b-tcomanr-merge-v2.3 --enable-reasoning --reasoning-parser deepseek_r1

Sglang

Run this command

python -m sglang.launch_server --model-path ertghiu256/Qwen3-4b-tcomanr-merge-v2.3 --reasoning-parser deepseek-r1

llama.cpp

Run this command

llama-server --hf-repo ertghiu256/Qwen3-4b-tcomanr-merge-v2.3

or

llama-cli --hf ertghiu256/Qwen3-4b-tcomanr-merge-v2.3

Ollama

View the model at ollama.com or Run this command

ollama run ertghiu256/Qwen3-4b-tcomanr-merge-v2.3:Q8_0

or for Q5_K_M quant

ollama run hf.co/ertghiu256/Qwen3-4b-tcomanr-merge-v2.3:Q5_K_M

or for IQ4_NL quant

ollama run hf.co/ertghiu256/Qwen3-4b-tcomanr-merge-v2.3:IQ4_NL

LM Studio

Search

ertghiu256/Qwen3-4b-tcomanr-merge-v2.3

in the lm studio model search list then download

Recomended parameters

temp: 0.6
num_ctx: ≥8192
top_p: 0.95
top_k: 10
Repeat Penalty: 1.1

Merge Method

This model was merged using the TIES merge method using Qwen/Qwen3-4B-Thinking-2507 as a base.

Models Merged

The following models were included in the merge:

• ertghiu256/qwen-3-4b-mixture-of-thought
• Tesslate/UIGEN-T3-4B-Preview-MAX
• ertghiu256/Qwen3-4B-Thinking-2507-Hermes-3
• ValiantLabs/Qwen3-4B-ShiningValiant3
• ertghiu256/qwen3-math-reasoner
• ValiantLabs/Qwen3-4B-Esper3
• Qwen/Qwen3-4b-Instruct-2507
• ertghiu256/qwen3-multi-reasoner
• janhq/Jan-v1-4B
• ertghiu256/qwen3-4b-code-reasoning
• ertghiu256/Qwen3-Hermes-4b
• GetSoloTech/Qwen3-Code-Reasoning-4B
• POLARIS-Project/Polaris-4B-Preview
• huihui-ai/Huihui-Qwen3-4B-Thinking-2507-abliterated

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: ertghiu256/qwen3-math-reasoner
    parameters:
      weight: 0.8
  - model: ertghiu256/qwen3-4b-code-reasoning
    parameters:
      weight: 0.9
  - model: ertghiu256/qwen-3-4b-mixture-of-thought
    parameters:
      weight: 1.0
  - model: POLARIS-Project/Polaris-4B-Preview
    parameters:
      weight: 0.8
  - model: ertghiu256/qwen3-multi-reasoner
    parameters:
      weight: 0.9
  - model: ertghiu256/Qwen3-Hermes-4b
    parameters:
      weight: 0.7
  - model: ValiantLabs/Qwen3-4B-Esper3
    parameters:
      weight: 0.75
  - model: Tesslate/UIGEN-T3-4B-Preview-MAX
    parameters:
      weight: 1.0
  - model: ValiantLabs/Qwen3-4B-ShiningValiant3
    parameters:
      weight: 0.6
      density: 0.5
  - model: huihui-ai/Huihui-Qwen3-4B-Thinking-2507-abliterated
    parameters:
      weight: 0.75
  - model: Qwen/Qwen3-4B-Thinking-2507
    parameters:
      weight: 1.0
  - model: Qwen/Qwen3-4b-Instruct-2507
    parameters:
      weight: 0.75
  - model: GetSoloTech/Qwen3-Code-Reasoning-4B
    parameters:
      weight: 0.75
      density: 0.55
  - model: ertghiu256/Qwen3-4B-Thinking-2507-Hermes-3
    parameters:
      weight: 1.0
  - model: janhq/Jan-v1-4B
    parameters:
      weight: 0.3
merge_method: ties
base_model: Qwen/Qwen3-4B-Thinking-2507
parameters:
  normalize: true
  int8_mask: true
  lambda: 1.0
dtype: float16