- Lower costs and 100+ new languages coming to Gemini 1.5 | Google Cloud Blog
- Google Cloud’s Vertex AI is rolling out multiple updates to make AI accessible for every enterprise including lower costs and more languages.
Google recently announced a price reduction for Gemini Flash overnight. The announcement stated that prices would be reduced by up to 85%.
I feel like I've written about 5 posts related to price reductions lately... It seems like prices are being reduced that much.
A few days ago, I posted aboutprice reduction, and now, just a few days later, there's another price reduction.
The details of the newly reduced prices are as follows:
The related Google document is as follows:
I've been posting about price reductions so much lately that I'm going to take a moment to organize everything and refrain from posting about prices for a while. I want to take some time to organize my thoughts before continuing.
Early last year, OpenAI's GPT-3 Davinci product was priced at a whopping $2 per 1M tokens for both Input and Output.
The newly reduced price for Google's Gemini Flash is $0.01875 for Input and $0.075 for Output. Based on the Input price, the difference is a staggering 100 times. In other words, it's dropped to 1/100th of the original price. (Furthermore, if you use Batch Prediction, this difference widens to 200 times.)
So, what about performance? The difference is so significant that it's almost embarrassing to compare. Flash is a leading LLM across various fields, (excluding the flagship model) and is inherently multilingual. In contrast, Davinci back then struggled even with Korean.
It's fascinating to see this rapid price competition and the resulting reductions in the industry.
Prices have been dropping almost by half every three months. When I started durumis last year, I was slightly concerned about translation costs, but even then, Google was already reducing prices. I thought, "Google will figure it out," and they did.
Now, at least in the LLM realm, we can almost completely disregard cost concerns and add features freely.
There's no guarantee that prices will continue to drop, but even at the current level, I believe we can implement almost anything we want to.
I hope you can share my excitement for the continued development of durumis...
Comments0