| Initial Release | 2023-02-24 |
|---|---|
| Based on | Base Model |
| Released by | Meta AI |
| Available Sizes | 7b, 13b, 33b, 65b |
| Announcement Paper | [[1]] |
| Huggingface Link | [[2]] |
LLaMA (Large Language Model Meta AI) is a foundational large language model (LLM) released by Meta AI in February 2023. It was originally designed primarily for use by researchers.
Four model sizes were trained: 7, 13, 33 and 65 billion parameters. LLaMA's developers reported that the 13B parameter model's performance on most NLP benchmarks exceeded that of the much larger GPT-3 (with 175B parameters) and that the largest model was competitive with state of the art models.
LLaMA was significant at the time of its release for the fact that its model weights were released to the research community under a noncommercial license, rather than being controlled behind an API. Within a week of LLaMA's release, its weights were leaked to the public on 4chan via BitTorrent.
Language Support[edit | edit source]
LLaMA primarily supports English. It should have limited capabilities in a range of other languages including: bg, ca, cs, da, de, es, fr, hr, hu, it, nl, pl, pt, ro, ru, sl, sr, sv, uk.
Derivative and Related Models[edit | edit source]
Due to general availability of its weights, LLaMA was the basis for a large number of significant derivative models - including:
A project is underway to replicate the LLaMA model in a pure Open Source way, see OpenLLaMA.
License[edit | edit source]
The LLaMA license restricts its usage to research and prohibits commercial use. For models distributed on more permissive terms see LLaMA 2 and OpenLLaMA.
See Also[edit | edit source]
- LLaMA model on Wikipedia