whatmodel.sh / GPT-J 6B

GPT-J 6B

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

GPT-J 6B is an open-weights language model from EleutherAI, released June 2021. Its published size is 6.05B parameters and a 2k-token context window. Licence: apache-2.0.

24.5±0.6#675 of 719 LLMsfrom 30 benchmark results across 6 sources

$ how it compares

The LLMs ranked either side of GPT-J 6B, out of 828 with a score, then more from EleutherAI.

#modelscorewhereRAMreleased
778RedPajama-INCITE Base 3B v1 Together25±3local~3.4GB2023-05
779ChatGLM2-6B Zhipu AI (THUDM)25±2local~15GB2023-06
780GPT-J 6B EleutherAI25±1local~14GB2021-06
781OpenLLaMA 3B v2 OpenLM Research25±1local~4.2GB2023-07
782OPT-6.7B Meta24±1local~16GB2022-05
more from EleutherAI
756GPT-NeoX-20B EleutherAI28±1local~14GB2022-02
777Pythia 12B EleutherAI25±1local~9.5GB2023-02
787Pythia 6.9B EleutherAI23±1local~6.1GB2023-02

see all LLMs · find the ones that fit your machine

$ what it's good at

coding▁▁▃▃▆██▇▆▄▃▁ #449 of 458 * · 28% of its score · 2 tests

Beats Molmo 7B-D (0924), DeepSeek R1 Distill Qwen 1.5B, Qwen3 0.6B here, though they rank higher overall. Trails Phi-1 (1.3B), OpenAssistant SFT-4 Pythia 12B, Falcon3 1B Instruct here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
HumanEval12.2% ✎▁·▃▁▃▄▆▅█▃▁▁ beats 3/99easy · avg 79%▇▇▇▇▇▇▇ 17%sharp
HumanEval+11.0% ✎▁·▁▃▁▃▄▄▄█▃▁ beats 3/99easy · avg 74%▇▇▇▇ 11%sharp
other tests15 not in a task · 72% of its score
testresultvs the others in its rundifficultycountswhy
ARC Challenge +136.5% ✎▁·▁··▁▁▇▃█▆▃ beats 11/26easy · avg 85%▇▇▇▇▇▇▇▇ 20%sharp · several runs
WinoGrande +264.2% ✎▁····▁··▁▂█▄ beats 9/26easy · avg 89%▇▇▇▇ 11%at its level · several runs
HellaSwag +266.2% ✎▃···▁▇▅█▇▃▁▁ beats 8/26very easy · avg 92%▇▇▇ 8.4%at its level · several runs
ARC Easy +162.2% ✎▁▁·▁▁▄▅▄█▅▁▃ beats 8/26easy · avg 88%▇▇ 5.1%at its level · several runs
SciQ +191.5% ✎▁·▁·▁·▃▅▃█▄▁ beats 10/26very easy · avg 98%▇▇ 4.6%several runs
BoolQ65.4% ✎▂▃▂▅▆▂▅▆█▂▂▂ beats 8/26easy · avg 91%▇▇ 4.3%at its level
PIQA +275.6% ✎▁·▁··▁▁▃█▅█▂ beats 8/26easy · avg 88%▇▇ 4.0%blunt · at its level · several runs
LAMBADA (OpenAI version), last-word accuracy +268.2% ✎▁··▂▁▁▁▂▆█▅▃ beats 11/26very easy · avg 90%▇▇ 4.0%blunt · at its level · several runs
7 more tests
testresultvs the others in its rundifficultycountswhy
TriviaQA +139.3% ✎▁··▁█▃▃▁▄▅▄█ beats 15/28easy · avg 72%▇ 3.2%several runs
Natural Questions (closed book)9.2% ✎▁·▁·▁▂▁▃·▅▂█ beats 8/28medium · avg 58%▇ 3.1%typical
OpenBookQA +129.0% ✎█·▁▁▃·▁▃▅▇▃▄ beats 12/26medium · avg 52%▇ 2.9%too hard for it · several runs
Social IQA41.1% ✎▁···▁▃█▁▁▁·▁ beats 6/26medium · avg 59%▇ 1.0%blunt
Open PT-LLM Leaderboard average (9 Portuguese tasks)35.7% ✎▃▁▁▃▁▁▁▁▂█▅▁ beats 43/276medium · avg 73%▇ 0.4%noisy
TruthfulQA MC236.0% ✎▅▃█▆▁▂▅···▁▁ beats 3/31medium · avg 64%▇ 0.3%blunt · at its level
HaluEval QA hallucination detection (accuracy)51.4% ✎█▁▁·▃▁·▁▁▁▁▁ beats 18/27medium · avg 70%▇ 0.1%sharp · too hard for it · noisy

$ where to run it

Other ways to run it

  1. Run it as published · needs ~12.11 GB · from huggingface.co/EleutherAI (official)
    uv run --with torch --with 'transformers<5' --with accelerate python -c "from transformers import pipeline; print(pipeline('text-generation', model='EleutherAI/gpt-j-6b', torch_dtype='auto', device_map='auto', revision='float16')('Once upon a time', max_new_tokens=40)[0]['generated_text'])"

Official weights and documentation: huggingface.co/EleutherAI ↗

$ how the score was worked out · Stability AI — original Stabl… 40%, EvalPlus / EvalPerf published… 28%, GPT-NeoX-20B: An Open-Source… 19% and 3 more

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
Stability AI — original StableLM zero-shot evaluation outputs10▇▇▇▇▇▇▇▇▇▇▇▇ 40%Apache-2.0
EvalPlus / EvalPerf published evaluation runs2▇▇▇▇▇▇▇▇ 28%Apache-2.0
GPT-NeoX-20B: An Open-Source Autoregressive Language Model — authors' own runs · lab-reported9▇▇▇▇▇▇ 19%CC BY 4.0
EleutherAI GPT-J-6B model card — EleutherAI lm-evaluation-harness runs · lab-reported4▇▇ 8%Apache-2.0
Hallucinations Leaderboard results (HF dataset hallucinations-leaderboard/results)4▇ 4%Apache-2.0
Open Portuguese LLM Leaderboard (archived)1▇ 0%Apache-2.0
prior · pull toward the middle▇ 0%
weighted average= 24.5

$ 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