whatmodel.sh / cont-PaLM 62B (PaLM 62B trained to 1.3T tokens)
cont-PaLM 62B (PaLM 62B trained to 1.3T tokens)
↗ share
+ contribute
cont-PaLM 62B (PaLM 62B trained to 1.3T tokens) 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 checkpoint trained further to ~1.3T tokens. The Flan paper (Table 2 caption) names it 'cont-PaLM (also known as PaLM 62B at 1.3T tokens)'; the PaLM paper v5 reports it as 'PaLM 62B' with 1325B training tokens (Tables 31-32, Table 40 one-shot column). Distinct from the original 795B-token palm-62b. Never released. Comparison-only.
Model information sources
Score within the LMArena (text) comparison set only.
$ how it compares
The LLMs ranked either side of cont-PaLM 62B (PaLM 62B trained to 1.3T tokens), out of 877 with a score, then more from Google.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 615 |  Ministral 3B (2410) | 43±1 | comparison only | ||
| 616 |  Phi-3-vision-128k-instruct | 43±2 | local | ||
| 617 | cont-PaLM 62B (PaLM 62B trained to 1.3T tokens) | 43±1 | never released | ||
| 618 |  DeepSeek LLM 67B Chat | 43±1 | local | ||
| 619 |  Qwen1.5 14B Chat | 43±1 | local | ||
| more from Google | |||||
| 11 |  Gemini 4 Argon | 80±1 | api $4.00/M | ||
| 26 | Gemini 3.8 Flash | 76±1 | api $1.50/M | ||
| 34 | Gemini 3.7 Flash | 75±1 | api $1.50/M | ||
$ what it's good at
coding█·█······█·█ #2 of 4 * · 14% of its score · 2 tests
math▁▂▃▆█▇▅▇▃▁▁▁ #284 of 523 * · 24% of its score · 2 tests
Beats GPT-4o (2024-08-06), Command A (03-2025), GPT-4o (2024-05-13) here, though they rank higher overall. Trails Gemma 7B (base), CodeGemma 7B (v1.0), Phi-2 (2.7B) here, though they rank lower overall.
other tests10 not in a task · 62% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| MMLU lab +1 | 61.2% ✎ | ▁···▁··▃▁▅▅█ beats 10/18 | 22% |
| BIG-Bench Hard lab | 41.7% ✎ | ▂▂▂▂·▅·▃▂█▃▃ beats 8/18 | 13% |
| TyDiQA-GoldP (1-shot, EM, 8-language average) lab | 45.7% ✎ | ▂··▂·▂▄▄·███ beats 10/18 | 5.8% |
| Natural Questions (closed book) lab | 21.8% ✎ | ▂▁▁▁▂▄▄▃▅█▅▁ beats 45/68 | 4.4% |
| WinoGrande lab | 77.0% ✎ | ▁▁▁▁▂▁▁▅▇█▃▁ beats 118/131 | 4.3% |
| TriviaQA lab | 71.8% ✎ | ▁▁▂▂▄▄▃▆▅█▅▂ beats 71/80 | 3.7% |
| BoolQ lab | 83.9% ✎ | ▁····▁▂█▆▅▆▄ beats 87/103 | 3.3% |
| HellaSwag lab | 80.6% ✎ | ▁▁▁▂▂▂▄▆█▇▄▂ beats 112/135 | 2.9% |
2 more tests
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LAMBADA (OpenAI version), last-word accuracy lab | 76.3% ✎ | ▁·····▁▁▃▃██ beats 81/86 | 1.3% |
| PIQA lab | 81.4% ✎ | ▁▁▁▂▂▃▅█▅▃▁▁ beats 105/123 | 1.2% |
$ where to run it
No verified API or download recorded yet · add one
$ how the score was worked out · PaLM: Scaling Language Modeli… 51%, Scaling Instruction-Finetuned… 48%
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
| source | results | share of the score |
|---|---|---|
 PaLM: Scaling Language Modeling with Pathways — authors' own runs · lab-reported | 11 | 51% |
| Scaling Instruction-Finetuned Language Models — authors' own runs · lab-reported | 4 | 48% |
| prior · pull toward the middle | 0% | |
| weighted average | = 43.1 |
$ improve this page
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