Google Gemini 1.5 - Google Shocked Everyone with 1,000,000 Token Context Window
ฝัง
- เผยแพร่เมื่อ 14 ก.พ. 2024
- Google just introduced a brand new AI model called Gemini 1.5 Pro.
The real power of Gemini 1.5 is its context window. It now has a maximum context window of 1 million tokens. However, most people will only get access to the 128000 context window.
A million tokens can be up to 700,000 words or up to 11 hours of video, 11 hours of audio And 30,000 thousands of line of code.
Besides the context window, Google is calming that Gemini 1.5 is now comparable to Ultra 1.0, which is the paid model.
Read the full blog post and watch the demo videos here: blog.google/technology/ai/goo...
![I Built a $1M AI App [No Code]](http://i.ytimg.com/vi/IZsQqarWXtY/mqdefault.jpg)
Google seems to announce a lot of very cool stuff that fails to impress when it’s released. I hope this time is different but I’m not holding my breathe.
Yep, you are right
I work for a company with an industry leading LLM-leveraging product. We've learned to ignore any of Google's claims before testing the model ourselves. It's a bit shameful on their end.
they're playing second fiddle to openai. It's good though as it means there's an AI war and that's why we're seeing so many crazy improvements so quickly.
its been shut down for racism and promoting pedophilia
very true
I tried asking Google advanced to read a document and it says it cannot, then I keep asking it to do and it for one random document in my Google drive .when I asked it to do for a second document in my goggle drive it says it can't read from Google drive when I asked how it did manage to summarise the first document it said it didn't or pretended to do it didn't actually read from my drive but when I opened the first document and checked it with gemini summaries it matched
You're not gonna be able to test Gemini Pro 1.5 on Bard/Gemini yet, only on AI Studio and/or Vertex AI if you're one of the few that get access to it through either or both of those platforms.
great quality of content
If a model supports 1 million tokens but can't actually process any text due to ridiculously strict content policy, is it really any good?
The Gemini system is an impeccable one for the AI development system that tends to give a greater performance I think specially with the power optimisation from mediatek tech
What video editing software do you use?
Adobe premiere mainly
I wonder how bad loss in the middle is? Definitely use rerank in your RAG apps when using this model
Are Gemini & Gemini Advanced both using the 1.5 model, and which one is the paid version?
Gemini itself refers to the family of large language models developed by Google AI, including various versions like 1.0 Pro and 1.5 (which comes in Pro and Flash flavors).
Gemini Advanced is a paid plan that grants access to a specific model within the Gemini family - Gemini 1.5 Pro. Along with 1.5 Pro access, it also offers additional features like document upload and analysis, and potentially more in the future.
So, both Gemini and Gemini Advanced utilize the 1.5 model, but Gemini Advanced is the paid tier that unlocks the 1.5 Pro version along with other benefits.
Great Information 👍🏻
I was under the impression this is all based off the vertex ai model
Can I upload my data?
Google is probably releasing this info to increase its stock price. Hype, as usual.
Thank you sir for the update.
These are my questions
1. How can I access Google studio? ( I want to play with the models and build tools, hope it's possible?)
2. Is Google API free or paid
I think it's only limited invite only. Haven't been able to find a way to test it yet.
so vile and racist its been shut down
I don't know what's with your voice in this video but it sounds disjoint between left and right. The discrepancy is really annoying when trying to listen to you.
In six months from now they will discontinue it leaving you hanging - it's what they do.