GPT-4.5 vs o3-mini

Performance Comparison

Factual Accuracy (SimpleQA)

GPT-4.5
62.5%
o3-mini
15%

Hallucination Rate (Lower is better)

GPT-4.5
37.1%
o3-mini
80.3%

Science Reasoning (GPQA)

GPT-4.5
71.4%
o3-mini
79.7%

Math Problem Solving (AIME)

GPT-4.5
36.7%
o3-mini
87.3%

GPT-4.5 vs o3-mini

Key Differences

Coding Performance (SWE-Lancer)

GPT-4.5
32.6%
o3-mini
10.8%

Multilingual (MMMLU)

GPT-4.5
85.1%
o3-mini
~60%

Best For:

GPT-4.5
General knowledge, factual accuracy, conversation
o3-mini
Complex reasoning, math, science problems

Training Focus:

GPT-4.5
SFT, RLHF, Scalable Alignment
o3-mini
Chain-of-thought (CoT) reasoning
VS