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

MATH-optimized | Aggressive pruning | 35% weights pruned

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

Note: Minimal quality drop detected. The Wanda pruning algorithm effectively identifies and removes less important weights while preserving model capability.

Performance Comparison

Category	Original	Pruned	Change
Python	0.0%	0.0%	→
Html	0.0%	0.0%	→
Trivia	35.0%	30.0%	↓ 5.0%
Math	15.0%	10.0% ⭐	↓ 5.0%
Reasoning	35.0%	30.0%	↓ 5.0%
Medical	50.0%	50.0%	→
Linux	5.0%	15.0%	↑ 10.0%
Writing	45.0%	40.0%	↓ 5.0%

Average: 23.1% -> 21.9% (-1.2%)

Math Retention: 66.7%

$Comparison Graph$

Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer

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

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))