DedeProGames/smallcoder

DedeProGames/ smallcoder

86 Downloads Updated 2 weeks ago

SmallCoder is a compact reasoning-focused coding model, fine-tuned from DeepSeek-R1 1.5B using a code dataset that includes step-by-step reasoning.

1.5b

ollama run DedeProGames/smallcoder:1.5b

curl http://localhost:11434/api/chat \
  -d '{
    "model": "DedeProGames/smallcoder:1.5b",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='DedeProGames/smallcoder:1.5b',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'DedeProGames/smallcoder:1.5b',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Models

Name

1 model

Size

Context

Input

smallcoder:1.5b

1.1GB · 128K context window · Text · 2 weeks ago

smallcoder:1.5b

1.1GB

128K

Text

Readme

Welcome to SmallCoder, a compact reasoning-focused coding model fine-tuned from DeepSeek-R1 1.5B using a code dataset that includes step-by-step reasoning.
It is designed to deliver strong coding performance for its parameter size, emphasizing structured reasoning and correctness during problem solving.

Overview

SmallCoder is a compact reasoning-focused coding model, fine-tuned from DeepSeek-R1 1.5B using a code dataset that includes step-by-step reasoning.
It targets strong code generation for its parameter size, prioritizing correctness-oriented reasoning during coding tasks.

The training approach focuses on exposing the model to structured, multi-step solutions so it can better handle logic-heavy programming tasks while remaining lightweight and efficient.

Benchmark Results

Model	LCB (v5) (8/1/24–2/1/25)	Codeforces Rating	Codeforces Percentile	HumanEval+
SmallCoder	25.1	963	28.5	73.0
Deepseek-R1-Distill-Qwen-1.5B	16.9	615	1.9	58.3

These results highlight improvements in reasoning-driven coding performance while maintaining a compact model size.

SmallCoder aims to balance efficiency and reasoning depth, making it suitable for local deployment, experimentation, and lightweight coding assistance workflows.

![upscalemedia-transformed (1).png](/assets/DedeProGames/smallcoder/5551ebe7-d81b-4f34-beac-919cf1256b73)

Welcome to **SmallCoder**, a compact reasoning-focused coding model fine-tuned from **DeepSeek-R1 1.5B** using a **code dataset** that includes step-by-step reasoning.  
It is designed to deliver strong coding performance for its parameter size, emphasizing structured reasoning and correctness during problem solving.

---

## Overview

**SmallCoder** is a compact reasoning-focused coding model, fine-tuned from **DeepSeek-R1 1.5B** using a **code dataset** that includes step-by-step reasoning.  
It targets strong code generation for its parameter size, prioritizing correctness-oriented reasoning during coding tasks.

The training approach focuses on exposing the model to structured, multi-step solutions so it can better handle logic-heavy programming tasks while remaining lightweight and efficient.

---

## Benchmark Results

| Model                              | LCB (v5) (8/1/24–2/1/25) | Codeforces Rating | Codeforces Percentile | HumanEval+ |
|------------------------------------|--------------------------|-------------------|-----------------------|------------|
| **SmallCoder**                    | 25.1                     | 963               | 28.5                  | 73.0       |
| Deepseek-R1-Distill-Qwen-1.5B      | 16.9                     | 615               | 1.9                   | 58.3       |

These results highlight improvements in reasoning-driven coding performance while maintaining a compact model size.

SmallCoder aims to balance efficiency and reasoning depth, making it suitable for local deployment, experimentation, and lightweight coding assistance workflows.

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)