LLaMA 3.1 vs GPT-5
Meta LLaMA 3.1 | OpenAI GPT-5 | |
|---|---|---|
| Overview | ||
| Company | Meta | OpenAI |
| Release date | Jul 23 2024 | Aug 7 2025 |
| Access | Open Weight | Proprietary |
| Benchmarks | ||
Nonsense detection BullshitBench v2Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better. | 14% | 21%Best |
Community preference (code) Arena Elo (Code)Like the text arena, but people vote on which AI writes better code. The votes become a chess-style Elo rating on arena.ai. Higher is better. | — | 1394 |
| Timeline | ||
| Release gap | LLaMA 3.1 shipped 380 days before GPT-5 | |
Which is better: LLaMA 3.1 or GPT-5?
GPT-5 leads LLaMA 3.1 on 1 of the 1 benchmark they both report (BullshitBench v2). LLaMA 3.1 shipped 380 days before GPT-5, so benchmark comparisons should account for the intervening progress.
LLaMA 3.1 is open weight, while GPT-5 is proprietary.
On BullshitBench v2, GPT-5 leads at 21% vs LLaMA 3.1 at 14%.
Frequently asked questions
LLaMA 3.1 was released by Meta on Jul 23 2024.
GPT-5 was released by OpenAI on Aug 7 2025.
LLaMA 3.1 is an open weight model released by Meta. GPT-5 is a proprietary model released by OpenAI.