Delphi

Closed weights Allen Institute for AI,University of Washington 11B parameters July 2022

No estimate

No hardware requirements for this model

The weights for this model have not been published, so it cannot be downloaded or run on your own hardware at any size. It is reachable only through its provider, and no graphics card changes that.

On record

Full specification

Everything on record for this model. Most of it describes how it was trained rather than how it runs — useful context for judging how much work went into it, and how it compares with models built at a different scale.

Origin

Who built this model, where, and when it was published.

Organisation
Allen Institute for AI,University of Washington
Organisation type
Research collective,Academia
Country
United States of America
Published
12 July 2022
Authors
Liwei Jiang, Jena D. Hwang, Chandra Bhagavatula, Ronan Le Bras, Jenny Liang, Jesse Dodge, Keisuke Sakaguchi, Maxwell Forbes, Jon Borchardt, Saadia Gabriel, Yulia Tsvetkov, Oren Etzioni, Maarten Sap, Regina Rini, Yejin Choi

What it does

The problem areas the model was built for. A model can carry several of each.

Domain
Language
Task
Chat
Base model
Unicorn

Size

How large the model is and how much data it was trained on. Parameters are the figure that decides whether it fits on a given graphics card.

Parameters
11B
Training data
26,214,400,000 tokens

Training on the proposed COMMONSENSE NORM BANK is carried out for 400k gradient updates

Epochs
4
Batch size
16

Training compute

The arithmetic performed to train the model, measured in floating-point operations. It is a measure of what the training run cost, not of how fast the finished model answers you.

How it was established
Operation counting
Fine-tuning compute
1.7 × 10¹⁸ FLOP

=6*11B*400k gradient updates*16 (batch size)*4 epochs=1.6896 × 10^18

The training run

What it physically took to train: which chips, how many, for how long, and what that drew from the wall.

Training hardware
Google TPU v3
Wall-clock time
72 hours

We train Delphi using TPU v3-32 and evaluate it using TPU v3-8, with model parallelisms of 32 and 8 respectively, on Google Cloud Virtual Machines. Training Delphi on COMMONSENSE NORM BANK for 4 epochs takes approximately 72 hours.

Availability

Whether you can obtain the model and run it on your own hardware, which is what decides if any of the graphics-card figures on this page apply.

Weights
Closed — provider access only
Model access
Hosted access (no API)
Training code
Unreleased

https://delphi.allenai.org/

How it is classified

Labels the source dataset applies when tracking notable models, and how confident it is in the entry.

Record confidence
Speculative

Sources

Where this record came from and when it was last checked.

Reference
Can Machines Learn Morality? The Delphi Experiment
Last updated
28 November 2025

What the numbers mean

About this model

Delphi was published by Allen Institute for AI,University of Washington, in United States of America, in July 2022. It comes out of research collective,Academia.

It works in Language, and is recorded as doing chat.

It is derived from Unicorn rather than trained from scratch, which is the usual way a specialised model is produced.

This is a closed model: the trained values stayed with whoever produced them, and there is no local version to run.

How it was trained

It was trained on about 26,214,400,000 tokens of text.

Answers

Delphi — common questions

01

What is Delphi used for?

Delphi works in Language, and is recorded as handling chat. A model can carry several of each, so these are the areas it was built for rather than a limit on what it will attempt.

02

What GPU do I need to run Delphi?

None. Delphi is a closed model — its weights were never published, so it cannot be downloaded or run on your own hardware at any price. It is reachable only through its provider.

03

Is Delphi open source?

No. Delphi has not had its weights published, so it exists only as a service controlled by its owner.

04

How many parameters does Delphi have?

Delphi has 11B parameters. That figure is the total, and it is what decides how much memory the model needs — roughly half a gigabyte per billion at the compression most people use.

05

Who created Delphi?

Delphi was published by Allen Institute for AI,University of Washington, based in United States of America, categorised as research collective,Academia.

06

When was Delphi released?

Delphi was published in July 2022. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.

Source

Original publication

Record last updated 28 November 2025

The other direction

Looking at it from the other side?

This page starts from the model. If you already own a card and want to know everything it will run, start from the hardware instead.