PaLM
Google DeepMind ·
Verified
3 of 3 published values traced to a source3 primary sources, 3 archived1 not researched yet
Why this status
Cited to the lab's own announcement and the technical paper. Not verified because the context window could not be found in any cited primary source — the value may be correct, but this dataset cannot yet show where it came from. Capabilities evidenced from the primary sources: multilingual: “PaLM was trained using a combination of English and multilingual datasets that include high-quality web documents, books, Wikipedia, conversations, and GitHub code.” | coding: “PaLM 540B shows strong performance across coding tasks and natural language tasks in a single model, even though it has only 5% code in the pre-training dataset.”.
What these states mean · How records are judged · Has this record changed?
Google's 540-billion-parameter model that demonstrated breakthrough few-shot scaling.
Details
| Released | April 4, 2022stated in research.googlePaLM was released on April 4, 2022.
|
|---|---|
| Company | Google DeepMind |
| Family | PaLM |
| Lineage | before PaLM 2 |
| Era | Pre-ChatGPT scaling |
| Type | Language model · LLM |
| Modalities | Text → Text |
| Capabilities | Coding, Multilingual |
| Weights | Proprietary |
| Licence | Not applicable — proprietary |
| Context window | 2K tokens |
| Parameters | 540B |
Current information, elsewhere
This record is a historical one: every figure above is traced to a dated primary source. For what is true today — live pricing, current rankings, availability — these are better maintained than anything here could be.
- Current price and providers OpenRouter ↗
- Current benchmark scores Artificial Analysis ↗
Sources
Record status: verified · confidence 90/100 · 3 of 3 primary
Cited to the lab's own announcement and the technical paper. Not verified because the context window could not be found in any cited primary source — the value may be correct, but this dataset cannot yet show where it came from. Capabilities evidenced from the primary sources: multilingual: “PaLM was trained using a combination of English and multilingual datasets that include high-quality web documents, books, Wikipedia, conversations, and GitHub code.” | coding: “PaLM 540B shows strong performance across coding tasks and natural language tasks in a single model, even though it has only 5% code in the pre-training dataset.”.
- research.google Official announcement archived
- arxiv.org Technical paper archived
- arxiv.org Technical paper archived
Primary means published by the organisation that made the model. Verified records require at least one.
PaLM family
- PaLM April 4, 2022
- PaLM 2 May 10, 2023
Compare PaLM with another model →
See PaLM on the interactive timeline →