BloombergGPT
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
- Bloomberg,Johns Hopkins University
- Organisation type
- Industry,Academia
- Country
- United States of America
- Published
- 30 March 2023
- Authors
- Shijie Wu, Ozan Irsoy, Steven Lu, Vadim Dabravolski, Mark Dredze, Sebastian Gehrmann, Prabhanjan Kambadur, David Rosenberg, Gideon Mann
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Language
- Task
- Language modeling, Language modeling/generation, Question answering, Financial management, Text classification
- Approach
- Self-supervised learning
- 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
- 50.6B
- Training data
- 569,000,000,000 tokens
- Epochs
- 0.8
- Batch size
- 4,200,000
708.9 billion tokens. At 0.75 English words per token, that's 532B words
"in the first 7,200 steps, we use a batch size of 1,024 (2.1M tokens), then switch to a batch size of 2,048 (4.2M tokens) for the remainder of training."
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
- 2.4 × 10²³ FLOP
- How it was established
- Reported,Hardware
2.36e23 per Table 4 (using our usual hardware method, 512 A100s over 53 days would be 512 * 312 teraFLOP/s * 53 * 24 * 3600 * 0.3 = 2.19e23)
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
- Chips used
- 512
- Chip-hours
- 650,240
- Wall-clock time
- 1,270 hours (52.9 days)
- Hardware utilisation
- MFU 24.6% · HFU 32.7%
- Power draw
- 408.3 kW
- Compute cost
- $369,586
"~53 days"
Table 4 indicates an average of 102e12 FLOP/sec per A100 during training. HFU = 102e12 / 312e12 = 0.327. Confirmed by calculating manually with number of steps and time per step. Note also: "Since we adopt activation checkpointing to reduce our memory footprint, this costs us an additional 0.33x TFLOPs per iteration due to repeated forward passes." Then, MFU = (102/1.33)/312 = 0.2460
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
- 1,299
"We validate BloombergGPT on standard LLM benchmarks, open financial benchmarks, and a suite of internal benchmarks that most accurately reflect our intended usage. Our mixed dataset training leads to a model that outperforms existing models on financial tasks by significant margins without sacrificing performance on general LLM benchmarks."
Sources
Where this record came from and when it was last checked.
- Reference
- BloombergGPT: A Large Language Model for Finance
- Last updated
- 25 May 2026
What the numbers mean
About this model
BloombergGPT was published by Bloomberg,Johns Hopkins University, in United States of America, in March 2023. industry,Academia is the category the publisher falls under.
It works in Language, and is recorded as doing language modeling, Language modeling/generation, Question answering, Financial management, Text classification.
This is a closed model: the trained values stayed with whoever produced them, and there is no local version to run.
Training and provenance
Producing it required around 2.4 × 10²³ FLOP of arithmetic, on NVIDIA A100, which is a statement about the training budget rather than about inference.
The training set ran to roughly 569,000,000,000 tokens.
Its inclusion criterion is sOTA improvement.
Answers
BloombergGPT — common questions
Is BloombergGPT open source?
No. BloombergGPT has not had its weights published, so it exists only as a service controlled by its owner.
How many parameters does BloombergGPT have?
BloombergGPT has 50.6B 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 BloombergGPT?
BloombergGPT was published by Bloomberg,Johns Hopkins University, based in United States of America, categorised as industry,Academia.
When was BloombergGPT released?
BloombergGPT was published in March 2023. 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 BloombergGPT used for?
BloombergGPT works in Language, and is recorded as handling language modeling, Language modeling/generation, Question answering, Financial management, Text classification. 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.
How much compute was used to train BloombergGPT?
Around 2.4 × 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 BloombergGPT?
None. BloombergGPT 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.
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.