INTRODUCING

GPT4All

Access to OpenAI's latest model, GPT-4, free of charge for the Harvard community.

Update May 2024: After serving hundreds of million tokens for Harvard students, researchers, and faculty, GPT4All has been sunset. If you are looking for similar functionality, please see OpenAI's current offerings through Harvard's ChatGPT Edu.

GPT-4 is by far the best available language model, substantially outperforming ChatGPT/GPT-3.5. The Harvard Computer Society AI Group is committed to sharing the state-of-the-art in artificial intelligence and machine learning with as much accessibility as possible. Toward this goal, we are providing free access to GPT-4 for members of the Harvard community. Users default to test access GPT-4 with a daily limit of 3 prompts per day. Anyone interested in full access to GPT-4, see below, open to all use cases.

  • Access anywhere. GPT4All utilizes OpenAI API to provide access to GPT-4. It functions identically to ChatGPT's text output. No taxing resources on your device.
  • Private. Inputs to GPT-4 go directly to OpenAI's servers. OpenAI has committed to not using any inputs from applications like GPT4All in their training.
  • Free of charge. This project is funded by AI@HCS. The cost of running GPT-4 is roughly 15x that of GPT-3.5, thus requiring some limits, but as few as possible! Prompts may be up to 350 words and GPT-4 outputs will be up to ~400 words.

Increased access?

We welcome all use cases that require increased limits. You can request further access via this form.

Notes: (1) GPT-4 will refer to itself as GPT-3. (2) Don't waste your prompts! (3) See issues? Contact us at ai@hcs.harvard.edu.