GPT-3 175B (davinci)
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
- OpenAI
- Organisation type
- Industry
- Country
- United States of America
- Published
- 28 May 2020
- Authors
- Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litwin, Scott Gray, Benjamin Chess, Jack Clark, C…
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Language
- Task
- Text autocompletion, Language modeling/generation
- 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
- 174.6B
- Training data
- 238,000,000,000 tokens
- Epochs
- 0.6
- Batch size
- 3,200,000
"we train GPT-3, an autoregressive language model with 175 billion parameters" Rather, it's 174.6 billion. Table D.1: 174,600 million parameters
From table 2.2, we determine that there are 410 + 19 + 12 + 55 + 3 = 499 billion tokens. We multiply this by 0.75 to give 374B words. 3.74e11 ======================== [Anson: I think the calculation below doesn't look at all the data, the CommonCrawl data only constitutes 60% of the data. Multiplying by 5/3 gives 4.75e11] "The CommonCrawl data was downloaded from 41 shards of monthly CommonCrawl covering 2016 to 2019, constituting 45TB of compressed plaintext before filtering and 570GB aft…
3.2M, per table 2.1
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
- 3.1 × 10²³ FLOP
- How it was established
- Reported
Table D.1 https://arxiv.org/abs/2005.14165
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
- 10,000
- Chip-hours
- 3,552,000
- Wall-clock time
- 355 hours (14.8 days)
- Hardware utilisation
- MFU 19.7%
- Power draw
- 5.1 MW
- Compute cost
- $2,116,866
- Cloud vendor
- Microsoft
14.8 days according to https://arxiv.org/ftp/arxiv/papers/2104/2104.10350.pdf
See table 4 of the carbon emissions paper: https://arxiv.org/pdf/2104.10350 MFU calculation from total compute: (6 FLOP/token/param * 174.6B params * 300B tokens) / (10k V100 * 125 TFLOPS/V100 * 14.8 days) = 0.1966 https://www.wolframalpha.com/input?i=%286*174.6+billion+*+300+billion%29+FLOP+%2F+%281250000+TFLOPS+*+14.8+days%29 HFU calculation from GPU throughput: The paper reports 24.6 TFLOPS throughput per V100, and reports the V100 peak performance of 125 TFLOPS (which is correct if trainin…
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
https://openai.com/blog/openai-api
How it is classified
Labels the source dataset applies when tracking notable models, and how confident it is in the entry.
- Frontier model
- Yes
- Foundation model
- Yes
- Likely above 10²³ FLOP
- Yes
- Why it is tracked
- Highly cited,Training cost
- Record confidence
- Confident
- Citations
- 58,240
- Benchmark data
- GPT-3 175B (davinci)
Sources
Where this record came from and when it was last checked.
- Reference
- Language Models are Few-Shot Learners
- Last updated
- 25 May 2026
What the numbers mean
About this model
GPT-3 175B (davinci) was published by OpenAI, in United States of America, in May 2020. industry is the category the publisher falls under.
It works in Language, and is recorded as doing text autocompletion, Language modeling/generation.
Its weights were never published, so it can only be reached through its provider. No graphics card changes that.
Training and provenance
Producing it required around 3.1 × 10²³ FLOP of arithmetic, on NVIDIA Tesla V100 DGXS 32 GB, which is a statement about the training budget rather than about inference.
Around 238,000,000,000 tokens went into training it.
It is tracked in the underlying dataset for one reason in particular: highly cited,Training cost.
Answers
GPT-3 175B (davinci) — common questions
How much compute was used to train GPT-3 175B (davinci)?
Around 3.1 × 10²³ 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 GPT-3 175B (davinci)?
None. GPT-3 175B (davinci) 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 GPT-3 175B (davinci) open source?
No. GPT-3 175B (davinci) has not had its weights published, so it exists only as a service controlled by its owner.
How many parameters does GPT-3 175B (davinci) have?
GPT-3 175B (davinci) has 174.6B parameters. "we train GPT-3, an autoregressive language model with 175 billion parameters" Rather, it's 174.6 billion. Table D.1: 174,600 million 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 GPT-3 175B (davinci)?
GPT-3 175B (davinci) was published by OpenAI, based in United States of America, categorised as industry.
When was GPT-3 175B (davinci) released?
GPT-3 175B (davinci) was published in May 2020. 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 GPT-3 175B (davinci) used for?
GPT-3 175B (davinci) works in Language, and is recorded as handling text autocompletion, Language modeling/generation. 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.