whatmodel.sh / GLaM dense 0.1B

GLaM dense 0.1B

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

GLaM dense 0.1B is a closed-weights language model from Google, released December 2021. Its published size is 0.13B parameters. Licence: proprietary.

Catalogue identity notes

Google's GLaM 0.1B dense baseline model (Du et al., arXiv:2112.06905, Table 4: 0.13B total / 0.13B activated parameters), trained on the GLaM dataset. Research size from the same paper. Never released. Not the same checkpoint as LaMDA-PT even where sizes match. Comparison-only.

Model information sources

Release date ↗

22.3±1.9comparison only: no current run option verifiedfrom 29 benchmark results across 1 source

Score within the ANLI Round 1 (prompted) comparison set only.

$ how it compares

Scored LLMs released closest to GLaM dense 0.1B, out of 1182 with a score, then more from Google.

#modelscorewhereRAMreleased
1034LaMDA 137B (pre-trained, LaMDA-PT) Google34±1never released2022-01
1035GPT-3 text-davinci-001 OpenAI34±1comparison only2022-01
1106LaMDA 68B (pre-trained) Google27±1never released2022-01
1156LaMDA 8B (pre-trained) Google12±2never released2022-01
1158GPT-3 text-curie-001 OpenAI12±1comparison only2022-01
more from Google
11Gemini 4 Argon Google80±1api $4.00/M2026-09
26Gemini 3.8 Flash Google76±1api $1.50/M2026-09
34Gemini 3.7 Flash Google75±1api $1.50/M2026-08

see all LLMs · find the ones that fit your machine

$ what it's good at

other tests29 not in a task · 99% of its score
testresultvs the others in its rundifficultycountswhy
COPA lab67.0% ✎▄·▄···▄·▄█·█ beats 0/7easy · avg 83%▇▇▇▇ 9.2%at its level
HellaSwag lab34.7% ✎▃··▃···▃·▃▃█ beats 0/7medium · avg 62%▇▇▇ 8.5%sharp
SQuAD 2.0 (prompted, F1) lab16.8% ✎██····██████ beats 0/7medium · avg 47%▇▇▇ 8.2%sharp
ReCoRD (SuperGLUE, prompted) lab71.6% ✎▄·▄·····▄▄██ beats 0/7very easy · avg 86%▇▇▇ 6.9%at its level
TriviaQA lab2.3% ✎▄··▄···▄·█▄█ beats 0/7hard · avg 35%▇▇▇ 6.7%sharp · too hard for it
Winograd Schema Challenge (WSC273) lab67.0% ✎█·····▄▄▄▄▄▄ beats 1/7medium · avg 79%▇▇▇ 6.5%at its level
SuperGLUE WSC (prompted) lab66.3% ✎█·····▄·▄█▄▄ beats 1/7medium · avg 77%▇▇ 5.8%at its level
Story Cloze Test lab63.3% ✎▄·▄····▄▄▄▄█ beats 0/7medium · avg 75%▇▇ 5.4%at its level
21 more tests
testresultvs the others in its rundifficultycountswhy
CoQA lab40.7% ✎▃▃·····█▃▃·▃ beats 0/7medium · avg 64%▇▇ 5.4%at its level
ARC Easy lab44.5% ✎▄··▄···▄·█▄█ beats 0/7medium · avg 63%▇▇ 4.4%at its level
RACE-middle lab40.6% ✎▄▄····▄▄·▄█▄ beats 0/7medium · avg 56%▇▇ 4.0%typical
LAMBADA (OpenAI version), last-word accuracy lab37.8% ✎▄▄·····▄·█▄█ beats 0/7medium · avg 60%▇▇ 3.9%at its level
PIQA lab64.4% ✎▃···▃··▃·▃█▃ beats 0/7medium · avg 75%▇ 3.4%at its level
RACE-high lab29.4% ✎▃▃······▅·▃█ beats 0/7hard · avg 39%▇ 2.9%too hard for it
WinoGrande lab49.7% ✎▃··▃·····▃▅█ beats 0/7hard · avg 61%▇ 2.6%sharp · too hard for it
Natural Questions (closed book) lab1.1% ✎▄·▄··▄·█▄·▄▄ beats 0/7very hard · avg 7.2%▇ 2.4%sharp · too hard for it
ARC Challenge lab23.2% ✎▃··▂·····▄·█ beats 0/7very hard · avg 35%▇ 2.4%sharp · too hard for it
OpenBookQA lab36.8% ✎▄·▄····█·█·█ beats 0/7hard · avg 46%▇ 2.4%typical
BoolQ lab56.6% ✎█···▃··▃▃·▃▃ beats 1/7hard · avg 65%▇ 2.2%sharp · too hard for it · noisy
WebQuestions (closed book) lab0.7% ✎▄····▄·▄█▄▄▄ beats 0/7very hard · avg 6.7%▇ 2.2%sharp · too hard for it
QuAC (prompted, F1) lab25.4% ✎█···▄·█▄··▄▄ beats 1/7hard · avg 32%▇ 1.2%blunt
DROP lab5.7% ✎▄·▄██▄·····▄ beats 0/7hard · avg 16%▇ 1.0%sharp · too hard for it · noisy
RTE (prompted) lab51.3% ✎█▄··▄····▄▄█ beats 3/7very hard · avg 55%▇ 0.4%sharp · too hard for it · noisy
ANLI Round 3 (prompted) lab34.8% ✎▄·▄▄██·····▄ beats 5/7very hard · avg 35%▇ 0.2%sharp · too hard for it · noisy
ANLI Round 1 (prompted) lab36.1% ✎█·▄▄·▄··▄··█ beats 5/7very hard · avg 35%▇ 0.2%too hard for it · noisy
MultiRC (SuperGLUE, prompted, F1a) lab58.6% ✎▃···▅▃█····▃ beats 6/7medium · avg 56%▇ 0.1%blunt · at its level · noisy
ANLI Round 2 (prompted) lab36.7% ✎█▄·····█·▄·█ beats 6/7very hard · avg 35%▇ 0.1%blunt · too hard for it · noisy
CommitmentBank (prompted) lab42.9% ✎█▄▄█·▄·····▄ beats 5/7medium · avg 43%▇ 0.0%blunt · at its level · noisy
WiC (prompted) lab50.8% ✎█▄·······▂·▃ beats 7/7very hard · avg 50%▇ 0.0%blunt · too hard for it · noisy

$ where to run it

No verified API or download recorded yet · add one

$ how the score was worked out · GLaM: Efficient Scaling of La… 99%

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
GLaM: Efficient Scaling of Language Models with Mixture-of-Experts — authors' own runs · lab-reported29CC BY 4.0
prior · pull toward the middle
weighted average= 22.3

$ 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