---
language: en
license: other
library_name: transformers
tags:
  - chat
  - reasoning
  - chain-of-thought
  - think
  - chichu
base_model: Sebastianpro88/Chichu-2.0-500M-Instruct
pipeline_tag: text-generation
---

# Chichu 2.5 Reasoning 🐱🧠

A 500M parameter language model with **chain-of-thought reasoning** using `<think>` blocks.
Fine-tuned from Chichu 2.0 on reasoning data covering math, logic, code, and general questions.

Named after Chichu the cat. 🐱

## What's New vs Chichu 2.0

Chichu 2.5 uses `<think>...</think>` blocks to reason through problems step by step before answering — similar to DeepSeek-R1 and Kimi-K3.

## Usage
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = AutoModelForCausalLM.from_pretrained("Sebastianpro88/Chichu-2.5-Reasoning", torch_dtype=torch.float16, device_map="cpu")
tokenizer = AutoTokenizer.from_pretrained("Sebastianpro88/Chichu-2.5-Reasoning")

messages = [
    {"role": "system", "content": "You are Chichu 2.5. Use <think> blocks to reason before answering."},
    {"role": "user", "content": "What is 15% of 240?"}
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt")
out = model.generate(**inputs, max_new_tokens=200, do_sample=False)
print(tokenizer.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))
```