Megatron-LM (8.3B)
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
- NVIDIA
- Organisation type
- Industry
- Country
- United States of America
- Published
- 17 September 2019
- Authors
- Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper, Bryan Catanzaro
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Language
- Task
- Language modeling/generation
- Approach
- Self-supervised learning
- Numerical format
- FP16
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
- 8.3B
- Training data
- 46,400,000,000 tokens
- Epochs
- 4.4
Source: https://lair.lighton.ai/akronomicon/ Archived source: https://web.archive.org/web/20211220142906/https://lair.lighton.ai/akronomicon/ Data also available on GitHub: https://github.com/lightonai/akronomicon/blob/main/akrodb/NVIDIA/Megatron-LM.json
"The resulting aggregate corpus contains 174 GB of deduplicated text."
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
- 9.1 × 10²¹ FLOP
- How it was established
- Hardware,Operation counting,Third-party estimation
source: https://lair.lighton.ai/akronomicon/ archived: https://github.com/lightonai/akronomicon/tree/main/akrodb other estimates: 8.3B is a GPT-2-based model (Table 2). "For GPT-2 models, all training is performed with sequences of 1024 subword units at a batch size of 512 for 300k iterations" I interpret the above as 1024*512*300k = 157B training tokens 6 * 157 billion * 8.3 billion = 7.8e21 Also, their training setup achieved 15.1 petaFLOPS or 1.5e16 FLOPS. (512 V100s is 512 * 125 ter…
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 Tesla V100 DGXS 32 GB
- Chips used
- 512
- Chip-hours
- 167,424
- Wall-clock time
- 327 hours (13.6 days)
- Hardware utilisation
- HFU 23.6%
- Power draw
- 262.6 kW
- Compute cost
- $109,288
Reported throughput is 15.1 teraFLOPS per GPU on 512 GPUs Assume total compute is 9.1e21 FLOP. Then training time is 327 hours. https://www.wolframalpha.com/input?i=9.1*10%5E21+FLOP+%2F+%28512*15.1+teraFLOPS%29
Achieved 15.1 PetaFLOP/s on 512 V100s, vs theoretical maximum of 512 * 125e12 = 64 PetaFLOP/s HFU = 15.1 / 64 = 0.2359
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
- Open source
code (2.5B model is a GPT model): https://github.com/NVIDIA/Megatron-LM?tab=readme-ov-file#megatron-overview open license: https://github.com/NVIDIA/Megatron-LM?tab=License-1-ov-file#readme
How it is classified
Labels the source dataset applies when tracking notable models, and how confident it is in the entry.
- Frontier model
- Yes
- Why it is tracked
- Highly cited,SOTA improvement
- Record confidence
- Likely
- Citations
- 2,766
- Benchmark data
- Megatron-LM (8.3B)
"Using the GPT-2 model we achieve SOTA results on the WikiText103 (10.8 compared to SOTA perplexity of 15.8) and LAMBADA" GPT-2 model here meaning model similar to GPT-2
Sources
Where this record came from and when it was last checked.
- Reference
- Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism
- Last updated
- 25 May 2026
What the numbers mean
Where it came from
Megatron-LM (8.3B) was published by NVIDIA, in United States of America, in September 2019. The organisation is categorised as industry.
It works in Language, and is recorded as doing language modeling/generation.
This is a closed model: the trained values stayed with whoever produced them, and there is no local version to run.
What went into building it
The training run consumed about 9.1 × 10²¹ FLOP, on NVIDIA Tesla V100 DGXS 32 GB. That figure describes the cost of creating it and has no bearing on how quickly it generates text.
It was trained on about 46,400,000,000 tokens of text.
Its inclusion criterion is highly cited,SOTA improvement.
Answers
Megatron-LM (8.3B) — common questions
When was Megatron-LM (8.3B) released?
Megatron-LM (8.3B) was published in September 2019. 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 Megatron-LM (8.3B) used for?
Megatron-LM (8.3B) works in Language, and is recorded as handling language modeling/generation. These are the areas it was designed around; they describe intent rather than a hard boundary.
How much compute was used to train Megatron-LM (8.3B)?
Around 9.1 × 10²¹ FLOP, on NVIDIA Tesla V100 DGXS 32 GB. 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 Megatron-LM (8.3B)?
None. Megatron-LM (8.3B) 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 Megatron-LM (8.3B) open source?
No. Megatron-LM (8.3B) has not had its weights published, so it exists only as a service controlled by its owner.
How many parameters does Megatron-LM (8.3B) have?
Megatron-LM (8.3B) has 8.3B parameters. Source: https://lair.lighton.ai/akronomicon/ Archived source: https://web.archive.org/web/20211220142906/https://lair.lighton.ai/akronomicon/ Data also available on GitHub: https://github.com/lightonai/akronomicon/blob/main/akrodb/NVIDIA/Megatron-LM.json. 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 Megatron-LM (8.3B)?
Megatron-LM (8.3B) was published by NVIDIA, based in United States of America, categorised as industry.
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.