whatmodel.sh / OPT-66B

OPT-66B

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

OPT-66B is an open-weights language model from Meta, released May 2022. Its published size is 66B parameters and a 2k-token context window. Licence: opt-175b-license-noncommercial.

Catalogue identity notes

Meta's Open Pre-trained Transformer base model as released on Hugging Face (weights under Meta's non-commercial OPT licence).

30.5±1.4#639 of 719 LLMsfrom 7 benchmark results across 2 sources

$ how it compares

The LLMs ranked either side of OPT-66B, out of 828 with a score, then more from Meta.

#modelscorewhereRAMreleased
734StarCoderBase 7B BigCode31±1local~6GB2023-07
735Llama 2 7B (base) Meta31±1local~5.5GB2023-07
736OPT-66B Meta31±1local~141GB2022-05
737LFM2 350M Liquid AI30±2local~1.7GB2025-07
738SmolLM2 1.7B Instruct Hugging Face30±1local~2.6GB2024-10
more from Meta
24Muse Spark 1.3 Meta77±1api $2.00/M2026-09
40Muse Spark 1.2 Meta75±1api $2.00/M2026-08
43Muse Spark 1.1 Meta75±1api $2.00/M2026-07

see all LLMs · find the ones that fit your machine

$ what it's good at

other tests7 not in a task · 99% of its score
testresultvs the others in its rundifficultycountswhy
ARC Challenge37.2% ✎█··▁·▁▁·▁▂▁▃ beats 19/19medium · avg 72%▇▇▇▇▇▇▇▇▇▇▇▇▇ 32%sharp
WinoGrande68.7% ✎▄····▄▆▂▄▆▆█ beats 19/19easy · avg 89%▇▇▇▇▇▇▇▇▇ 22%at its level
ARC Easy71.7% ✎▃·▅▃▃▅·██▃██ beats 19/19very easy · avg 90%▇▇▇▇▇▇▇ 18%at its level
PIQA78.8% ✎▂·▄▄·▄▂▆▂▄█▄ beats 19/19easy · avg 88%▇▇▇▇ 10%blunt · at its level
SciQ92.6% ✎▂··▂▃▂▂▆▃▃█▃ beats 19/19very easy · avg 98%▇▇▇ 7.6%typical
LAMBADA (OpenAI version), last-word accuracy73.9% ✎▃··▅▃▃▅▅▅███ beats 19/19very easy · avg 89%▇▇▇ 7.1%blunt · at its level
Open PT-LLM Leaderboard average (9 Portuguese tasks)35.0% ✎▃▁▁▃▁▁▁▁▂█▅▁ beats 41/276medium · avg 73%▇ 2.6%noisy

$ where to run it

Other ways to run it

  1. Run it as published · needs ~132.37 GB · from huggingface.co/facebook
    uv run --with torch --with 'transformers<5' --with accelerate python -c "from transformers import pipeline; print(pipeline('text-generation', model='facebook/opt-66b', torch_dtype='auto', device_map='auto')('Once upon a time', max_new_tokens=40)[0]['generated_text'])"

Weights and documentation: huggingface.co/facebook ↗

$ how the score was worked out · Pythia: A Suite for Analyzing… 97%, Open Portuguese LLM Leaderboa… 3%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling — authors' own runs6▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 97%CC BY-SA 4.0
Open Portuguese LLM Leaderboard (archived)1▇ 3%Apache-2.0
prior · pull toward the middle▇ 1%
weighted average= 30.5

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json