whatmodel.sh / GLaM 0.1B/64E (MoE)

GLaM 0.1B/64E (MoE)

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

GLaM 0.1B/64E (MoE) is a closed-weights language model from Google, released December 2021. Its published size is 1.9B parameters (0.145B active per token). Licence: proprietary.

Catalogue identity notes

Google's GLaM 0.1B/64E mixture-of-experts model (Du et al., arXiv:2112.06905, Table 4: 1.9B total / 0.145B 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 ↗

30.1±1.7comparison 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 0.1B/64E (MoE), 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
TriviaQA lab9.4% ✎▄··▄···▄·█▄█ beats 1/7hard · avg 35%▇▇▇▇ 10%sharp
HellaSwag lab43.1% ✎▃··▃···▃·▃▃█ beats 1/7medium · avg 62%▇▇▇▇ 9.0%sharp
SQuAD 2.0 (prompted, F1) lab22.9% ✎██····██████ beats 1/7medium · avg 47%▇▇▇ 8.5%sharp
COPA lab73.0% ✎▄·▄···▄·▄█·█ beats 1/7easy · avg 83%▇▇▇ 8.0%at its level
Winograd Schema Challenge (WSC273) lab66.3% ✎█·····▄▄▄▄▄▄ beats 0/7medium · avg 79%▇▇ 5.8%at its level
SuperGLUE WSC (prompted) lab65.6% ✎█·····▄·▄█▄▄ beats 0/7medium · avg 77%▇▇ 5.2%at its level
ReCoRD (SuperGLUE, prompted) lab77.5% ✎▄·▄·····▄▄██ beats 1/7very easy · avg 86%▇▇ 5.1%typical
Story Cloze Test lab66.4% ✎▄·▄····▄▄▄▄█ beats 1/7medium · avg 75%▇▇ 4.9%at its level
21 more tests
testresultvs the others in its rundifficultycountswhy
CoQA lab45.9% ✎▃▃·····█▃▃·▃ beats 1/7medium · avg 64%▇▇ 4.5%at its level
ARC Easy lab52.0% ✎▄··▄···▄·█▄█ beats 1/7medium · avg 63%▇▇ 4.0%at its level
RACE-middle lab43.4% ✎▄▄····▄▄·▄█▄ beats 1/7medium · avg 56%▇▇ 3.8%typical
WinoGrande lab51.0% ✎▃··▃·····▃▅█ beats 1/7hard · avg 61%▇ 3.6%sharp · too hard for it
Natural Questions (closed book) lab2.2% ✎▄·▄··▄·█▄·▄▄ beats 1/7very hard · avg 7.2%▇ 3.4%sharp · too hard for it
LAMBADA (OpenAI version), last-word accuracy lab41.4% ✎▄▄·····▄·█▄█ beats 1/7medium · avg 60%▇ 3.3%at its level
ARC Challenge lab26.5% ✎▂··▃·····▄·█ beats 1/7very hard · avg 35%▇ 3.3%sharp · too hard for it
RACE-high lab30.4% ✎▃▃······▅·▃█ beats 1/7hard · avg 39%▇ 3.2%typical
PIQA lab70.0% ✎▃···▃··▃·▃█▃ beats 1/7medium · avg 75%▇ 2.9%at its level
WebQuestions (closed book) lab3.4% ✎▄····▄·▄█▄▄▄ beats 1/7very hard · avg 6.7%▇ 2.7%sharp · too hard for it
BoolQ lab56.6% ✎█···▃··▃▃·▃▃ beats 2/7hard · avg 65%▇ 2.7%sharp · noisy
OpenBookQA lab40.0% ✎▄·▄····█·█·█ beats 1/7hard · avg 46%▇ 2.2%typical
DROP lab9.4% ✎▄·▄██▄·····▄ beats 1/7hard · avg 16%▇ 1.2%sharp · too hard for it · noisy
QuAC (prompted, F1) lab25.2% ✎█···▄·█▄··▄▄ beats 0/7hard · avg 32%▇ 1.1%blunt
RTE (prompted) lab45.8% ✎█▄··▄····▄▄█ beats 1/7very hard · avg 55%▇ 0.5%sharp · too hard for it · noisy
ANLI Round 3 (prompted) lab32.8% ✎▄·▄▄██·····▄ beats 0/7very hard · avg 35%▇ 0.3%sharp · too hard for it · noisy
ANLI Round 1 (prompted) lab34.1% ✎█·▄▄·▄··▄··█ beats 3/7very hard · avg 35%▇ 0.2%too hard for it · noisy
MultiRC (SuperGLUE, prompted, F1a) lab57.7% ✎▃···▅▃█····▃ beats 4/7medium · avg 56%▇ 0.1%blunt · at its level · noisy
ANLI Round 2 (prompted) lab33.8% ✎█▄·····█·▄·█ beats 2/7very hard · avg 35%▇ 0.1%blunt · too hard for it · noisy
CommitmentBank (prompted) lab66.1% ✎█▄▄█·▄·····▄ beats 7/7medium · avg 43%▇ 0.0%blunt · at its level · noisy
WiC (prompted) lab50.0% ✎█▄·······▂·▂ beats 3/7very 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= 30.1

$ 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