GPT Neox 20B — Benchmarks

Benchmark scores for GPT Neox 20B aggregated from public leaderboards (6 boards), with how it ranks among open models and how it compares with its nearest neighbours. See hardware requirements for what you need to run it.

No llmrun Composite Score for GPT Neox 20B: the score is only published for a model measured on at least 4 of the boards in the score panel, spanning at least 2 benchmark categories and including a reasoning or coding board — or on 3 boards that narrow its score to a 90% range of at most 15.6 points. Either way the results must come from at least two different benchmarks. Its individual scores below stand on their own — see how the llmrun Composite Score gets calculated.

GPT Neox 20B reasoning benchmarks

Multi-step logic and problem-solving — can the model think through a hard problem rather than recall an answer.

WinoGrande
WinoGrande is a large-scale commonsense-reasoning test built around pronoun resolution: the model must pick which of two nouns a pronoun refers to in a sentence, using real-world knowledge to disambiguate. It was designed by AI2 to resist the shortcuts that let earlier models game its predecessor, Winograd.
66.1%
66.1%
median 70.1%open leader 89.2%
40th percentile · #31 of 51 open · #58 of 79 all
#31 of 51
#58 of 79
Lab-reported
Epoch AI · Apr 2022

Other results (1)· leaderboard-only, not part of the llmrun Score

PIQA
PIQA (Physical Interaction QA) asks a model to choose the more sensible of two ways to accomplish an everyday physical task, testing practical, physical-world commonsense that isn't well captured by purely textual reasoning tests.
76.7%
76.7%
median 80.1%open leader 88.6%
Overall best: GPT 4o Mini (Jul 18, 2024) 88.7%
28th percentile · #29 of 40 open · #47 of 59 all
#29 of 40
#47 of 59
Lab-reported
Epoch AI · Apr 2022

GPT Neox 20B knowledge benchmarks

Breadth and depth of factual knowledge across many subjects, paired with the reasoning to apply it.

ARC Challenge
ARC Challenge is a set of grade-school science questions, curated by the Allen Institute for AI (AI2) to exclude ones that simple word-matching or retrieval can answer. It measures science knowledge paired with the reasoning needed to apply it.
41.1%
41.1%
median 53.9%open leader 95.3%
27th percentile · #41 of 56 open · #60 of 76 all
#41 of 56
#60 of 76
Lab-reported
Epoch AI · Apr 2022

Other results (3)· leaderboard-only, not part of the llmrun Score

HellaSwag
HellaSwag is a commonsense benchmark that asks a model to pick the most plausible continuation of an everyday situation. The wrong options are adversarially chosen to fool models, testing grounded commonsense understanding.
70.5%
70.5%
median 76.2%open leader 89.2%
Overall best: GPT 4 (Mar 14) 95.3%
28th percentile · #34 of 47 open · #55 of 75 all
#34 of 47
#55 of 75
Lab-reported
Epoch AI · Apr 2022
BoolQ
BoolQ is a reading-comprehension benchmark of naturally occurring yes/no questions, each paired with a short passage that contains the answer. It tests whether a model can extract and reason over information from real text rather than just recall facts.
64.9%
64.9%
median 76.4%open leader 89.4%
Overall best: T5 11B 91.2%
16th percentile · #32 of 38 open · #64 of 76 all
#32 of 38
#64 of 76
Lab-reported
Epoch AI · Apr 2022
OpenBookQA
OpenBookQA is a science-question benchmark modeled on open-book exams: the model is given a small set of core science facts and must combine them with broader common knowledge to answer multi-step questions. It targets reasoning over recall.
38.8%
38.8%
median 56.8%open leader 88.0%
24th percentile · #20 of 26 open · #34 of 41 all
#20 of 26
#34 of 41
Lab-reported
Epoch AI · Apr 2022

Scores aggregated from public benchmark sources (Epoch AI: 6 boards) — each linked from the benchmark pages. llmrun does not run these benchmarks. Percentiles and ranks are among open models in the llmrun catalogue; the median and leader marks refer to the same field.