The market for artificial intelligence is evolving - and OpenAI reacts with a price reduction. GPT-3 becomes significantly cheaper.
The market for cloud AI models is developing rapidly: In May 2020, OpenAI introduced the large language model GPT-3, and a month later the company announced paid access via the OpenAI API.
As of November 2021, OpenAI is now waiving the already short waiting period for new customers and making GPT-3 available to all. In the summer of the same year, the first competitors appeared, such as the Jurassic-1 Jumbo model from the Israeli AI startup AI21 Labs and alternatives for the Chinese market like Wu Dao 2.0.
OpenAI GPT-3 has real competition
2022 added more alternatives with the open-source language models GPT-NeoX-20B from EleutherAI, OPT-66B from Meta and the BLOOM from the BigScience project released in July.
The 176 billion parameter BLOOM model has recently been available for use via Eteria.ai. Currently, there is still a waiting list there.
Eteria promises competitive pricing and, according to its statement, can offer access to BLOOM at a lower price than the competition thanks to the efficient use of infrastructure and resources.
OpenAI significantly lowers prices for OpenAI API
For example, until recently, processing 1,000 tokens at OpenAI for the largest model (Davinci) cost $0.06; at Eteria, customers get 1,000 tokens for $0.04.
Now, after almost two years, OpenAI has announced a price reduction for all GPT-3 models. According to this, 1,000 tokens for Davinci will only cost 0.02 US dollars in the future, which means that the price will be cut in three. The same price reduction also applies to the Embeddings models, which enable more complex text processing.
| Model | Previously | From 01 September |
| Davinci | $0.06 / 1k tokens | $0.02 / 1k token |
| Curie | $0.006 / 1k token | $0.002 / 1k token |
| Babbage | $0.0012 / 1k token | $0.0005 / 1k token |
| Ada | $0.0008 / 1k token | $0.0004 / 1k tokens |
| Davinci Embeddings | $0.6 / 1k token | $0.2 / 1k tokens |
| Curie Embeddings | $0.06 / 1k token | $0.02 / 1k Token |
| Babbage Embeddings | $0.012 / 1k Token | $0.005 / 1k Token |
| Ada Embeddings | $0.008 / 1k tokens | $0.004 / 1k token |
"We have been looking forward to this change for a long time. Our teams have made incredible progress in making our models more efficient to run," OpenAI said. That has significantly reduced deployment costs, according to the company.
The price reduction for all models listed is effective the first of September. Fine-tuned GPT-3 models with additional training will not benefit from the price reduction.