There's an ongoing arms race in the artificial intelligence sphere, particularly between big players like OpenAI, Microsoft, and Google. These three companies are locked into a never-ending battle of advancement, and Google is especially an outsider given that it doesn't use OpenAI's GPT technology at all. Google instead has been using its own home-grown technology, initially starting with PaLM and Bard and now more recently switching to Gemini 1.0 and Gemini 1.5. With Gemini 1.5 Pro already seemingly doing the impossible, we've never been more excited for Gemini 1.5 Ultra.

To be clear, there is no confirmed launch period for Gemini 1.5 Ultra. Google announced its Gemini 1.5 model in February, with the first available model being the Pro model. We imagine that Ultra is currently being built and tested, given that Google said that 1.5 Pro was the "first" model from 1.5 family to be launched. One thing is clear though: Gemini 1.5 Ultra will likely need to compete with GPT-5, not GPT-4.

Gemini 1.0 Ultra already outpaces GPT-4

Not in every way, though

Gemini 1.0 Ultra has a few aces up its sleeve when it comes to outpacing GPT-4, scoring higher in a number of key benchmarks like MMLU. It's also one of the best in the world at programming, again beating out GPT-4. These are all thi9ngs that Google is already ahead in, and with 1.5 Ultra presumably on the horizon, it's expected that Google will further its lead.

However, GPT-4 still has a few key advantages. including when it comes to common sense reasoning and everyday tasks. Plus, using Gemini Advanced, you can't generate code that will run in the same way that you can with ChatGPT's GPT-4. GPT-4 in ChatGPT will actually run the code for you, generate graphs and other analytical supplements that you may need.

All of this is to say that Gemini 1.0 and GPT-4 are quite close, so any further advancement is very likely to give Google the edge.

Gemini 1.5 Ultra could do so much more

Google's context window is key

The most interesting part about Gemini 1.5 is the extended context window that Google announced that it supports. While it will support up to one million tokens in general usage at a higher cost, the company said that it had tested it with a context window of 10 million tokens. For reference, a token is roughly half a word, and that equates to a huge amount of text or other input. Furthermore, OpenAI's GPT-4 Turbo only has a context window of 128,000 tokens.

If Gemini 1.5 Ultra can benefit from that extended context window, complete with Ultra's capabilities, that's a massive step forward that would put Google far in the lead when it comes to AI. Its context window is already ahead of the competition and enables it to do significantly more, particularly in multi-modal capacities.

In other words, whatever Google does with Gemini 1.5 Ultra will instantly have an advantage because of the extended context window. That is, assuming it makes its way to the higher-tier LLM too. There's no reason to believe it won't, certainly not a larger context window in general, anyway. Google's Gemini context window is the ace up its sleeve, and something that others haven't got anywhere close to, comparatively.

Looking to the future of GPT-5 and Google Gemini 1.5 Ultra

GPT-5 may come this year

With pressure mounting against OpenAI, it's entirely possible that GPT-5 could arrive later this year. The company's CEO, Sam Altman, has been talking about GPT-5 more and more in recent months, all but confirming that it's currently being worked on. While we don't know what it's capable of, it's expected that the company is feeling the rest of the industry starting to catch up with it. With rumors that it has also achieved AGI internally, there are a lot of eyes on OpenAI right now.

If Google wants to truly compete, then it's not just short-term gains that the company needs to chase. Beating GPT-4 is great, but what it launches needs to compete with the next generation of artificial intelligence from competitors, too. That's why Gemini 1.5 Ultra is so exciting, as it may be our first glimpse into what the future of AI will really be capable of.