LFM2.5-1.2B-Thinking-math-safe

🎯 MATH-optimized | 📦 Safe pruning | ⚡ 1% weights pruned

This model is a conservatively pruned version of LiquidAI/LFM2.5-1.2B-Thinking.

Performance Comparison

Category	Original	Pruned	Change
Python	12.5%	2.5%	↓ 10.0%
Html	22.5%	20.0%	↓ 2.5%
Trivia	82.5%	75.0%	↓ 7.5%
Math	95.0%	95.0% ⭐	→
Reasoning	80.0%	82.5%	↑ 2.5%
Medical	92.5%	92.5%	→
Linux	82.5%	82.5%	→
Writing	70.0%	67.5%	↓ 2.5%

Average: 67.2% → 64.7% (-2.5%)

Math Retention: 100.0%

$Comparison Graph$

Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("CompactAI/LFM2.5-1.2B-Thinking-math-safe")
tokenizer = AutoTokenizer.from_pretrained("CompactAI/LFM2.5-1.2B-Thinking-math-safe")

inputs = tokenizer("Your prompt here", return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=100)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))