whatmodel.sh / U-PaLM 62B

U-PaLM 62B

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

U-PaLM 62B is a closed-weights language model from Google, released October 2022. Its published size is 62.5B parameters. Licence: proprietary.

Catalogue identity notes

Google's PaLM 62B continued with the UL2R objective (Tay et al., arXiv:2210.11399, Tables 2-3). Never released. Comparison-only.

Model information sources

Release date ↗

42.3±2.1comparison only: no current run option verifiedfrom 7 benchmark results across 1 source

Score within the LMArena (text) comparison set only.

$ how it compares

The LLMs ranked either side of U-PaLM 62B, out of 877 with a score, then more from Google.

#modelscorewhereRAMreleased
628Mixtral 8x7B Instruct v0.1 Mistral42±0local~28GB2023-12
629Nous Hermes 2 Mistral 7B DPO Nous Research42±3local~6.1GB2024-02
630U-PaLM 62B Google42±2never released2022-10
631Granite 4.0 1B IBM42±1local~2.4GB2025-10
632OpenHermes-2.5-Code-290k-13B ajibawa-202342±1local~9.8GB2024-03
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 tests7 not in a task · 98% of its score
testresultvs the others in its rundifficultycountswhy
WinoGrande lab76.2% ✎▁▁▁▁▂▁▁▅▇█▃▁ beats 115/131easy · avg 85%▇▇▇▇▇▇▇▇ 20%at its level
TriviaQA lab68.3% ✎▁▁▂▂▄▄▃▆▅█▅▂ beats 65/80easy · avg 79%▇▇▇▇▇▇▇ 17%at its level
HellaSwag lab79.7% ✎▁▁▁▂▂▂▄▆█▇▄▂ beats 108/135easy · avg 87%▇▇▇▇▇▇ 16%at its level
BoolQ lab85.4% ✎▁····▁▂█▆▅▆▄ beats 94/103easy · avg 90%▇▇▇▇▇▇ 16%at its level
Natural Questions (closed book) lab18.7% ✎▂▁▁▁▂▄▄▃▅█▅▁ beats 41/68hard · avg 29%▇▇▇▇▇▇ 16%typical
LAMBADA (OpenAI version), last-word accuracy lab79.7% ✎▁·····▁▁▃▃██ beats 85/86easy · avg 84%▇▇▇ 7.2%blunt
PIQA lab81.4% ✎▁▁▁▂▂▃▅█▅▃▁▁ beats 104/123easy · avg 85%▇▇ 5.6%blunt · at its level

$ where to run it

No verified API or download recorded yet · add one

$ how the score was worked out · Transcending Scaling Laws wit… 98%

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
Transcending Scaling Laws with 0.1% Extra Compute — maker's own results · lab-reported7Factual vendor-reported score (not licensed content)
prior · pull toward the middle
weighted average= 42.3

$ 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