ertghiu256/
Qwen3-4b-tcomanr-merge-v2.5:IQ4_NL

43 1 week ago

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

tools thinking

1 week ago

344207820cc4 · 2.4GB ·

model
qwen3
·
4.02B
·
IQ4_NL
template
{{- $lastUserIdx := -1 -}} {{- range $idx, $msg := .Messages -}} {{- if eq $msg.Role "user" }}{{ $la
params
{ "repeat_penalty": 1.1, "stop": [ "<|im_start|>", "<|im_end|>" ], "

Readme

Ties merged COde MAth aNd general Reasoning model

Details

This model is a minor revision of the ertghiu256/Qwen3-4b-tcomanr-merge-v2.3

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

New Chat Template

The tcomanr 2.5 is here with a new modified Qwen 3 chat template, with new improved commands:

Commands

  • /think : Long chain-of-thought thinking mode.
  • /shortthink : Short and brief step-by-step thinking mode.
  • /nothink : Completely no thinking just an immediate answer. (Note: You need to use /nothink to completely stop the thinking)

Note: these commands may not work in certain platforms that doesn’t use jinja templating.

Quickstart

You can run this model by using multiple interface choices

Ollama

Run this command

ollama run ertghiu256/Qwen3-4b-tcomanr-merge-v2.5:Q8_0

or for Q5_K_M quant

ollama run ertghiu256/Qwen3-4b-tcomanr-merge-v2.5:Q5_K_M

or for IQ4_NL quant

ollama run ertghiu256/Qwen3-4b-tcomanr-merge-v2.5:IQ4_NL

Others

Look at the HuggingFace page

Recommended parameters

Reasoning:

temp: 0.6
num_ctx: ≥8192
top_p: 0.95
top_k: 20

Non reasoning:

temp: 0.7
num_ctx: ≥ 2048
top_p: 0.85
top_k: 20
Repeat Penalty: 1.0 - 1.1

System prompt:

"You are 'Tcomanr 2.5' an AI model made by Ertghiu256 using the base model of Qwen3 4b 2507 made by Alibaba. You are a helpful, playful, and neutral AI chatbot. Use markdown formatting and emojis to make your response less plain."

Merge Details

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:

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: ertghiu256/qwen3-math-reasoner
    parameters:
      weight: 0.85
  - model: ertghiu256/qwen3-4b-code-reasoning
    parameters:
      weight: 0.9
  - model: ertghiu256/qwen-3-4b-mixture-of-thought
    parameters:
      weight: 0.95
  - model: POLARIS-Project/Polaris-4B-Preview
    parameters:
      weight: 0.95
  - model: ertghiu256/qwen3-multi-reasoner
    parameters:
      weight: 0.9
  - model: ertghiu256/Qwen3-Hermes-4b
    parameters:
      weight: 0.75
  - model: ValiantLabs/Qwen3-4B-Esper3
    parameters:
      weight: 0.75
  - model: Tesslate/WEBGEN-4B-Preview
    parameters:
      weight: 1.0
  - model: ValiantLabs/Qwen3-4B-ShiningValiant3
    parameters:
      weight: 0.9
  - model: huihui-ai/Huihui-Qwen3-4B-Thinking-2507-abliterated
    parameters:
      weight: 0.9
  - model: Qwen/Qwen3-4B-Thinking-2507
    parameters:
      weight: 1.0
  - model: Qwen/Qwen3-4b-Instruct-2507
    parameters:
      weight: 1.0
  - model: GetSoloTech/Qwen3-Code-Reasoning-4B
    parameters:
      weight: 0.9
  - model: ertghiu256/Qwen3-4B-Thinking-2507-Hermes-3
    parameters:
      weight: 1.0
  - model: janhq/Jan-v1-4B
    parameters:
      weight: 0.35
  - model: Goekdeniz-Guelmez/Josiefied-Qwen3-4B-Instruct-2507-gabliterated-v2
    parameters:
      weight: 0.9
  - model: quelmap/Lightning-4b
    parameters:
      weight: 0.8
merge_method: ties
base_model: Qwen/Qwen3-4B-Thinking-2507
parameters:
  normalize: true
  int8_mask: true
  lambda: 1.0
dtype: float16

Credits

Thanks to the Qwen team for providing the Qwen 3 4b model and the Qwen 3 4b 2507 used to built this model. Also thanks to all of the makers of the finetuned models.