GPT-4.5 vs Llama 3.1 70B (Instruct)
Compare GPT-4.5 by OpenAI against Llama 3.1 70B (Instruct) by Meta AI, context windows of 128K vs 128K, tested across 23 shared challenges.
Our Verdict
GPT-4.5
Llama 3.1 70B (Instruct)Runner-up
No community votes yet. On paper, GPT-4.5 has the edge — bigger model tier, newer.
Llama 3.1 70B (Instruct) is 190x cheaper per token — worth considering if cost matters.
Too close to call
API pricing
Cost per 1M tokens
GPT-4.5
Input
$75.00
Output
$150.00
Llama 3.1 70B (Instruct)
Input
$0.59
127× cheaper
Output
$0.79
190× cheaper
Llama 3.1 70B (Instruct) is cheaper on both — 127× input, 190× output
Writing DNA
Style Comparison
Similarity
47%
GPT-4.5 uses 129.6x more headings
GPT-4.5
Llama 3.1 70B (Instruct)
63%Vocabulary51%
16wSentence Length18w
0.57Hedging0.46
4.7Bold3.4
6.4Lists5.6
0.00Emoji0.00
1.30Headings0.00
0.54Transitions0.06
Based on 11 + 15 text responses
Research
What we learned reading every model
FAQ
