Gemini 1.5 could be ~20x cheaper than GPT4
3 min. read
Published on
Share this article
Improve this guide
Read our disclosure page to find out how can you help MSPoweruser sustain the editorial team Read more
Key notes
- Google’s Gemini 1.5 could be 20 times cheaper than OpenAI’s GPT-4.
- This is due to Gemini’s efficient architecture that reduces computational costs.
Google’s upcoming large language model, Gemini 1.5, is becoming a major challenge to OpenAI’s dominance in the API market. Early benchmarks suggest that Gemini 1.5 is competitive with OpenAI’s GPT-4 in terms of capabilities while offering a significant price advantage.
A Redditor who did the math believes Gemini 1.5’s pricing could be around 20 times lower than GPT-4. This significant cost difference stems from Gemini’s highly efficient architecture, which utilizes a technique called “soft Mixture of Experts” (MoE). This allows the model to activate only a small subset of its components at any given time, significantly reducing computational requirements and costs.
With Gemini 1.5 offering arguably superior performance at a fraction of the cost, OpenAI faces mounting pressure to respond. If they fail to develop a more efficient and competitive model soon, Google could potentially capture a significant portion of the API market, which is OpenAI’s primary source of revenue.
According to the post:
We can reasonably expect that pricing for 1.5 Pro should be similar to 1.0 Pro. Pricing for 1.0 Pro is $0.000125 / 1K characters.
—
Compare that to $0.01 / 1K tokens for GPT4-Turbo. Rule of thumb is about 4 characters / token, so that’s $0.0005 for 1.5 Pro vs $0.01 for GPT-4, or a 20x difference in Gemini’s favor.
—
So Google will be providing a model that is arguably superior to GPT4 overall at a price similar to GPT-3.5.
—
If OpenAI isn’t able to respond with a better and/or more efficient model soon Google will own the API market, and that is OpenAI’s main revenue stream.
The competition between Google and OpenAI is expected to continue driving innovation in the field of large language models. This ultimately benefits users with access to increasingly powerful and affordable AI tools.
Users had different takes on it; one had a Neutral stance, and they said,
“I am not a fanboy of either company. I’ll take whoever gives me a better product. That said, Google can put up or shut up. Both OpenAI and Google are posturing, but until we have a public product in our hands not under ‘laboratory conditions’, it’s just bluster and smoke. We all saw how disappointing Gemini 1.0 was.”
Meanwhile, the other looked quite positive,
“This is exciting because I think the competition will push OpenAI harder. OpenAI hasn’t had any real competition in about a year.”
One questioned the take that Gemini 1. is cheaper than GPT4, and they said,
It’s important to note that these are user opinions and not official statements from Google or OpenAI. The actual pricing and capabilities of Gemini 1.5 remain to be seen when it is publicly released.
