GLaM
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
- Organisation type
- Industry
- Country
- United States of America
- Published
- 13 December 2021
- Authors
- Nan Du, Yanping Huang, Andrew M. Dai, Simon Tong, Dmitry Lepikhin, Yuanzhong Xu, Maxim Krikun, Yanqi Zhou, Adams Wei Yu, Orhan Firat, Barret Zoph, Liam Fedus, Maarten Bosma, Zongwei Zhou, Tao Wang, Yu Emma Wang, Kellie Webster, Marie Pellat, Kevin Robinson, Kathleen Meier-Hellstern, Toju Duke, Lucas Dixon, Kun Zhang, Quoc V Le, Yonghui Wu, Zhifeng Chen, Claire Cui
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Language
- Task
- Language modeling/generation, Question answering
- Numerical format
- BF16
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
- 1.2T
- Training data
- 600,000,000,000 tokens
- Batch size
- 1,000,000
1.2 trillion parameters
The dataset is made of 1.6 trillion tokens, but later in the paper they say they only train the largest model for 600b tokens. 600b / 0.75 words/token = 800b words. "The complete GLaM training using 600B tokens consumes only 456 MWh and emits 40.2 net tCO2e."
"We use a maximum sequence length of 1024 tokens, and pack each input example to have up to 1 million tokens per batch."
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
- 3.6 × 10²³ FLOP
- How it was established
- Operation counting,Hardware
The network activates 96.6 billion parameters per token and trained for 600B tokens. 6 * 600B * 96.6B = 3.478e23 Digitizing figure 4 (d) indicates 139.67 TPU-years of training. 2.75e14 * 139.67 * 365.25 * 24 * 3600 * 0.3 = 3.636e23 Since these are close, we will use the 6NC estimate and derive hardware utilization from the training time information. Later they say they measured 326W power usage per chip, which could maybe be used to estimate utilization.
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 v4
- Chips used
- 1,024
- Chip-hours
- 1,398,784
- Wall-clock time
- 1,366 hours (56.9 days)
- Hardware utilisation
- MFU 28.7%
- Power draw
- 701.5 kW
- Compute cost
- $541,437
Note that they give several energy estimates. Use the complete training figures for 600B tokens, not the GPT-3 comparison values with 280B tokens. "326W measured system power per TPU-v4 chip" "The complete GLaM training using 600B tokens consumes only 456 MWh" 1024 TPU v4 chips (456 MWh) / (326W/chip * 1024 chips) = 1366 hours
6ND: 6 * 600B * 96.6B = 3.478e23 Hardware: Digitizing figure 4 (d) indicates 139.67 TPU-years of training. 2.75e14 * 139.67 * 365.25 * 24 * 3600 = 1.212e24 at full utilization Implies utilization of 0.287 MFU = 0.2870
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
- Unreleased
- Training code
- Unreleased
How it is classified
Labels the source dataset applies when tracking notable models, and how confident it is in the entry.
- Foundation model
- Yes
- Likely above 10²³ FLOP
- Yes
- Why it is tracked
- SOTA improvement
- Record confidence
- Confident
- Citations
- 1,198
"As shown in Table 5, GLaM (64B/64E) is better than the dense model and outperforms the previous finetuned state-of-the-art (SOTA) on this dataset in the open-domain setting"
Sources
Where this record came from and when it was last checked.
- Reference
- GLaM: Efficient Scaling of Language Models with Mixture-of-Experts
- Last updated
- 25 May 2026
What the numbers mean
Background
GLaM was published by Google, in United States of America, in December 2021. It comes out of industry.
It works in Language, and is recorded as doing language modeling/generation, Question answering.
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
Producing it required around 3.6 × 10²³ FLOP of arithmetic, on Google TPU v4, which is a statement about the training budget rather than about inference.
It was trained on about 600,000,000,000 tokens of text.
It is tracked in the underlying dataset for one reason in particular: sOTA improvement.
Answers
GLaM — common questions
How much compute was used to train GLaM?
Around 3.6 × 10²³ FLOP, on Google TPU v4. 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 GLaM?
None. GLaM 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 GLaM open source?
No. GLaM has not had its weights published, so it exists only as a service controlled by its owner.
How many parameters does GLaM have?
GLaM has 1.2T parameters. 1.2 trillion 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.
Who created GLaM?
GLaM was published by Google, based in United States of America, categorised as industry.
When was GLaM released?
GLaM was published in December 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 GLaM used for?
GLaM works in Language, and is recorded as handling language modeling/generation, Question answering. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.
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.