AI Training Costs

The cluster focuses on the high financial and computational costs of training large AI models like GPT-3 and GPT-4, with debates on exact figures, feasibility for hobbyists versus corporations, and trends in decreasing expenses.

➡️ Stable 1.0x AI & Machine Learning
3,123
Comments
15
Years Active
5
Top Authors
#3820
Topic ID

Activity Over Time

2011
1
2012
1
2014
1
2015
2
2016
27
2017
27
2018
49
2019
83
2020
185
2021
103
2022
210
2023
801
2024
610
2025
968
2026
55

Keywords

LLM techcrunch.com K2 AI GPT4 OAI i.e OP SD MSFT training train models model cost gpt inference openai 000 million

Sample Comments

Barrin92 Sep 12, 2021 View on HN

fully training a ~170 billion parameter model on the cheapest cloud instance probably sets you back at least a few million bucks so unless OP is Tony Stark that might not be as trivial as you think

p1esk Sep 23, 2020 View on HN

It costs ~$4.5M to train GPT-3. How big is your group of hobbyists?

mbrumlow Aug 8, 2019 View on HN

Are they even factoring in the cost of compute required to train their systems ?

JanSt Sep 27, 2024 View on HN

Yes, but 1) you only need to train the model once and the inference is way cheaper. Train one great model (i.e. Claude 3.5) and you can get much more than $80/month worth out of it. 2) the hardware is getting much better and prices will fall drastically once there is a bit of a saturation of the market or another company starts putting out hardware that can compete with NVIDIA

ugh123 Oct 13, 2023 View on HN

High training times could be cost prohibitive. Currently, its over $100mil to train GPT4 from scratch (which possibly includes other costs related to RLHF and data acquisition). Not sure how this model compares, but its likely not cheap.

hedora Sep 9, 2025 View on HN

Source? Large model training runs cost more than flying a plane across the atlantic, so this doesn’t sound right.

gjs278 Nov 8, 2021 View on HN

why not? people are spending 6k on cloud costs to train models in this thread. it pays for itself after two models.

bart_spoon Dec 16, 2023 View on HN

It’s the cost of compute hardware required to train a model of that size

vimy Apr 8, 2023 View on HN

Maybe. The problem is that you need billions to train new models.[1] At least with how things are now.[1] https://techcrunch.com/2023/04/06/anthropics-5b-4-year-plan-...

weinzierl Jun 29, 2023 View on HN

That sounds cheap. Can I really train a 13b model from scratch for just USD 2000?