GPT-J Models by EleutherAI
Last refreshed 2026-10-12. Next refresh: weekly.
Details
About
GPT-J is an open-source, autoregressive language model developed by EleutherAI, notable for its most prominent variant, GPT-J-6B, which features 6 billion parameters. This model stands as one of the largest publicly available in the GPT-3 style category, utilizing the transformer architecture. It has been trained on extensive datasets like the Pile, designed to generate human-like text responses from given prompts. While it excels in various natural language processing tasks, its core strength is text generation. The open accessibility of GPT-J's code and weights encourages research and development, although it requires responsible use due to potential biases or offensive content it may generate 458.
Decision facts
Current Variants
Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.
Use when the workload needs 2K context and 6B parameters.
| Model | Use when | Released | Signals | Status |
|---|---|---|---|---|
| GPT-J-6B | Use when the workload needs 2K context and 6B parameters. | Unknown release | 2K context6B parameters | Current |
Release Timeline
1 release groupSpecifications(1 models)
| Model | Released | Context | Parameters |
|---|---|---|---|
| GPT-J-6B | — | 2K | 6B |
