Jurassic-1-Jumbo
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
- AI21 Labs
- Organisation type
- Industry
- Country
- Israel
- Published
- 11 August 2021
- Authors
- Opher Lieber, Or Sharir, Barak Lenz, Yoav Shoham
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Language
- Task
- Language modeling/generation, Chat
- Approach
- Self-supervised learning
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
- 178B
- Training data
- 300,000,000,000 tokens
- Batch size
- 3,200,000
"Jurassic-1 models come in two sizes, where the Jumbo version, at 178B parameters, is the largest and most sophisticated language model ever released for general use by developers."
"Our model was trained with the conventional self-supervised auto-regressive training objective on 300B tokens drawn from publicly available resources" 1 token ~ 0.75 words
"Namely, we used a base learning rate of 1.2 × 10−4 and 0.6 × 10−4 , and a batch size of 2M and 3.2M tokens, for J1-Large and J1-Jumbo, respectively. We also used a linear warm-up over roughly the first 375 million tokens, and gradually increased the batch size from 32K tokens up to its target value for the first few billion tokens."
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.
- Training compute
- 3.7 × 10²³ FLOP
- How it was established
- Third-party estimation
see here https://docs.google.com/document/d/1B8x6XYcmB1u6Tmq3VcbAtj5bzhDaj2TcIPyK6Wpupx4/edit 6 * 178B * 300B = 3.204000e+23
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
- NVIDIA A100
- Compute cost
- $836,700
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
- API access
- Training code
- Unreleased
How it is classified
Labels the source dataset applies when tracking notable models, and how confident it is in the entry.
- Frontier model
- Yes
- Likely above 10²³ FLOP
- Yes
- Why it is tracked
- Training cost
- Record confidence
- Confident
- Citations
- 55
"Training such a large model, on over 800 GPUs over many months" Lower-bound cost estimate: 800 GPUs * $1/GPU-hour * 4 months = $2.3M True cost was probably substantially higher. "many months" implies more than 4, and the GPUs were probably more expensive than $1/hour.
Sources
Where this record came from and when it was last checked.
- Reference
- Jurassic-1: Technical Details and Evaluation
- Last updated
- 28 November 2025
What the numbers mean
What this model is
Jurassic-1-Jumbo was published by AI21 Labs, in Israel, in August 2021. industry is the category the publisher falls under.
It works in Language, and is recorded as doing language modeling/generation, Chat.
Because the weights are not available, none of the hardware figures elsewhere on this site apply to it.
What went into building it
Producing it required around 3.7 × 10²³ FLOP of arithmetic, on NVIDIA A100, which is a statement about the training budget rather than about inference.
It was trained on about 300,000,000,000 tokens of text.
It is tracked in the underlying dataset for one reason in particular: training cost.
Answers
Jurassic-1-Jumbo — common questions
How much compute was used to train Jurassic-1-Jumbo?
Around 3.7 × 10²³ FLOP, on NVIDIA A100. That measures what producing the model cost and says nothing about how quickly it answers once trained — inference speed comes from memory bandwidth, not from the training budget.
What GPU do I need to run Jurassic-1-Jumbo?
None. Jurassic-1-Jumbo 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.
Is Jurassic-1-Jumbo open source?
No. Jurassic-1-Jumbo has not had its weights published, so it exists only as a service controlled by its owner.
How many parameters does Jurassic-1-Jumbo have?
Jurassic-1-Jumbo has 178B parameters. "Jurassic-1 models come in two sizes, where the Jumbo version, at 178B parameters, is the largest and most sophisticated language model ever released for general use by developers.". 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.
Who created Jurassic-1-Jumbo?
Jurassic-1-Jumbo was published by AI21 Labs, based in Israel, categorised as industry.
When was Jurassic-1-Jumbo released?
Jurassic-1-Jumbo was published in August 2021. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.
What is Jurassic-1-Jumbo used for?
Jurassic-1-Jumbo works in Language, and is recorded as handling language modeling/generation, 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.
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.