whatmodel.sh / LaMDA 68B (pre-trained)

LaMDA 68B (pre-trained)

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

LaMDA 68B (pre-trained) is a closed-weights language model from Google, released January 2022. Its published size is 68B parameters. Licence: proprietary.

Catalogue identity notes

Smaller pre-trained LaMDA model evaluated by Google in Wei et al. (arXiv:2201.11903v6: 'LaMDA ... which has models of 422M, 2B, 8B, 68B, and 137B parameters'; table label '68B'). Released date is the LaMDA paper month (the LaMDA paper itself describes sizes 2B-137B). Never released. Comparison-only.

Model information sources

Release date ↗

26.8±0.8comparison only: no current run option verifiedfrom 9 benchmark results across 1 source

Score within the LMArena (text) comparison set only.

$ how it compares

The LLMs ranked either side of LaMDA 68B (pre-trained), out of 877 with a score, then more from Google.

#modelscorewhereRAMreleased
805ChatGLM2-6B Zhipu AI (THUDM)28±2local~15GB2023-06
806RWKV-4-Raven 14B RWKV (BlinkDL)28±2comparison only2023-04
807LaMDA 68B (pre-trained) Google27±1never released2022-01
808BLOOMZ 3B BigScience27±2comparison only2022-11
809Qwen1.5 1.8B Chat Alibaba26±2local~2.7GB2024-01
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

math▁▂▃▆█▇▅▇▃▁▁▁ #468 of 523 * · 9.1% of its score · 1 test

Beats Yi 34B Chat, Llama 3 8B Instruct, Command R (03-2024) here, though they rank higher overall. Trails GPT-Neo 125M, GPT-2 XL (1.5B), GPT-Neo 1.3B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
GSM8K lab8.2% ✎█·▅▃▃▃·▃·▃▃▅ beats 7/12easy · avg 67%▇▇▇▇ 9.1%sharp · too hard for it
other tests8 not in a task · 91% of its score
testresultvs the others in its rundifficultycountswhy
MAWPS (CoT-paper union) lab30.6% ✎▅▃█···▅·▃▃▃▅ beats 7/12very easy · avg 94%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 37%sharp · at its level
SVAMP lab18.8% ✎▃▅█···▅·▃▃▃▅ beats 7/12easy · avg 82%▇▇▇▇▇▇▇ 18%sharp
ASDiv lab23.1% ✎▅▃▃▅··▃▃·▃▃█ beats 6/12easy · avg 83%▇▇▇▇▇▇▇ 17%sharp
CommonsenseQA lab37.0% ✎█······▃·▁▁▃ beats 7/12easy · avg 87%▇▇▇ 6.8%sharp
BIG-bench Date Understanding lab15.5% ✎▃·▅·▅▃▅·▃▃·█ beats 7/12medium · avg 52%▇▇ 4.1%typical
BIG-bench Sports Understanding lab55.2% ✎█·····▃·▁·▄▁ beats 7/12medium · avg 81%▇ 3.6%sharp
AQuA-RAT lab20.2% ✎█▁·▁▁·····▂▁ beats 0/12medium · avg 48%▇ 1.8%sharp · too hard for it
StrategyQA lab59.6% ✎█·····▅▃▃▅▃█ beats 8/12medium · avg 70%▇ 1.5%typical

$ where to run it

No verified API or download recorded yet · add one

$ how the score was worked out · Chain-of-Thought Prompting El… 100%

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
Chain-of-Thought Prompting Elicits Reasoning in Large Language Models — authors' own runs · lab-reported9CC BY 4.0
prior · pull toward the middle
weighted average= 26.8

$ 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