LLaMA 3.1 vs o1
Meta LLaMA 3.1 | OpenAI o1 | |
|---|---|---|
| Overview | ||
| Company | Meta | OpenAI |
| Release date | Jul 23 2024 | Dec 5 2024 |
| 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% | — |
| Timeline | ||
| Release gap | LLaMA 3.1 shipped 135 days before o1 | |
Which is better: LLaMA 3.1 or o1?
LLaMA 3.1 and o1 don't publish scores on any of the same benchmarks, so there's no direct head-to-head comparison. LLaMA 3.1 shipped 135 days before o1, so benchmark comparisons should account for the intervening progress.
LLaMA 3.1 is open weight, while o1 is proprietary.
Direct benchmark comparisons are unavailable — LLaMA 3.1 and o1 don't publish scores on any of the same benchmarks.
Frequently asked questions
LLaMA 3.1 was released by Meta on Jul 23 2024.
o1 was released by OpenAI on Dec 5 2024.
LLaMA 3.1 is an open weight model released by Meta. o1 is a proprietary model released by OpenAI.