Jump to content

OpenAI Spring Update Event 2024 - GPT-4o Becomes Free to Use!

LAwLz

In an attempt to one-up and steal some thunder from Google IO that starts tomorrow, OpenAI has decided to hold an event today which they call their "OpenAI Spring Update".

 

 

 

 

Rumors and speculation:

Spoiler

Updates to ChatGPT and GPT-4. Maybe GPT-4.5?

Possibly an announcement about them working with Apple?

Maybe a search engine (although Sam seems to have said that it won't be announced today).

News about Sora, OpenAIs generative video creator?

A digital assistant similar to the one from the movie "her". With a lifelike voice that is able to understand speech.

Maybe OpenAI will make GPT-4 the default, free model?

One person who works at OpenAI tweeted that what they will announce would be cooler than GPT-5, but then quickly deleted the tweet.

 

 

 

You can watch the recorded event here:

 

More info can be found on OpenAI's blog:

Hello GPT-4o

Introducing GPT-4o and more tools to ChatGPT free users

 

 

 

 

 

Desktop version of ChatGPT (MacOS only, Windows version coming "later this year")

This seems to mostly be an Electron app. Basically, it's like the browser version, but looks and feels slightly more native than running it in a browser.

 

Edit: Never mind, it does have some additional features, such as having a shortcut to "copy" data into ChatGPT directly. So if you copy some code inside your IDE using a special keyboard shortcut, the code will be sent to ChatGPT and you can then ask it questions related to the code.

 

The desktop app also has the vision capability, so you can share your entire screen to ChatGPT. It can see what you see.

 

 

 

An update to the Web UI.

Honestly, it doesn't look like much changed. A very minor update in my opinion:

image.thumb.png.45c8b77d4a388d7b4727b301403ab886.png

 

 

GPT-4o is the latest and best model from OpenAI - And it's free for everyone!

The latest GPT model is called GPT-4o (four oh). The o stands for "omnimodel".

 

It seems like GPT-4o will be about the same in terms of output quality as GPT-4 Turbo, but with a few extra benefits.

Edit: Never mind, GPT-4o is also quite a bit better than GPT-4 Turbo. A while ago, a mysterious model appeared on the LMSys Arena website. As it turns out, the mysteriously named "gpt2" was GPT-4o.

Here are the results for the "arena battle" that was conducted during that limited time:

 

GPT-4o = 1310 points

GPT-4 Turbo = 1253 points

Gemini 1.5 Pro = 1249 points

Claude 3 Opus = 1246

 

Image:

Spoiler

Overallquestions.thumb.jpg.993fdf54b5bca423d6f9a4cb100abe4b.jpg

 

For the more advanced prompts (like coding or things harder than just writing a response to "what's up"), the ELO lead becomes even larger:

 

GPT-4o = 1369 points

GPT-4 Turbo = 1269 points

Claude 3 Opus = 1253 points

Gemini 1.5 Pro = 1238 points

 

 

Image:

Spoiler

Harderquestions.thumb.jpg.77e6b126aaf31fcd55f8015622e0f608.jpg

 

 

 

 

It seems like the custom GPTs that people publish will also become available to everyone now since you used to need GPT-4 for that before.

 

The "memory" support is also available for free now. For those who don't know, the plus subscription lets you add information about you which gets saved and used in all conversations. For example if I added "I live in Sweden" to the memory, ChatGPT will know and remember that in all future chats. So if I say "what should I do during the summer", it will suggest things that are suitable to do for a Swedish person during the summer. If I tell it I like playing video games it will use that information in conversations. 

 

ChatGPT can also, for free now, analyze images and uploaded documents. 

 

These last few things I have mentioned is already available to plus subscribers, now it will now be available for free to everyone.

 

 

Faster Reponses

GPT-4o is faster than GPT4 Turbo. According to OpenAI, the new model is twice as fast at generating output.

 

 

Better Voice Support

With GPT-4o, OpenAI has upgraded its voice capabilities.

It is way faster, so it start responding to your a lot quicker.

You can interrupt it while it is speaking back and give it other commands, and it will interrupt itself and adapt to your new command.

The generative voice has a much larger range of emotions and ways it can talk back to you. You can tell it to sound like a robot, sound emotional, and so on.

The model can also interpret your emotions and will adapt its response base on that.

 

 

Camera Support

In the live demo, OpenAI took out a phone, turned on the camera, and then asked ChatGPT to help them understand how to solve a linear calculation step by step.

ChatGPT analyzed the video feed from the camera and then gave voice directions on how to solve the equation step by step, like a math tutor.

They also showed it a paper with the words "I ❤️ ChatGPT" on it, and ChatGPT responded to it.

 

 

 

GPT-4o API update

OpenAI is updating its APIs.

 

The GPT-4o API will be 2x faster than GPT-4 Turbo.

It will be half the price of the GPT-4 Turbo API.

It will have 5x higher rate limit compared to GPT-4 Turbo.

 

Yes, that's right. OpenAI is cutting their prices in half once again.

 

 

Why would You Subscribe to The Plus Plan Now?

It seems like the biggest difference between the free and Plus version is that Plus subscribers will have 5x higher rate limit. What does this mean? We don't know yet. It might be that free users will be limited to for example 20 generated responses an hour, while Plus subscribers gets 100. The actual numbers remain to be seen.

Link to comment
Share on other sites

Link to post
Share on other sites

1 hour ago, LAwLz said:

 

One person who works at OpenAI tweeted that what they will announce would be cooler than GPT-5, but then quickly deleted the tweet.

 

Unless it's a holodeck, I'm not interested.

 

edit: so they gave it a sense of humor.

Link to comment
Share on other sites

Link to post
Share on other sites

6 hours ago, Kisai said:

Unless it's a holodeck, I'm not interested.

 

edit: so they gave it a sense of humor.

Nah, the projection will make enough money.
 

 

Link to comment
Share on other sites

Link to post
Share on other sites

It's still very crude and slow, also mess up. Need that Bladerunner hilo gf though. Also, far from Hal 9000 heh. Just don't make programming contradiction oof.

| Ryzen 7 7800X3D | AM5 B650 Aorus Elite AX | G.Skill Trident Z5 Neo RGB DDR5 32GB 6000MHz C30 | Sapphire PULSE Radeon RX 7900 XTX | Samsung 990 PRO 1TB with heatsink | Arctic Liquid Freezer II 360 | Seasonic Focus GX-850 | Lian Li Lanccool III | Mousepad: Skypad 3.0 XL / Zowie GTF-X | Mouse: Zowie S1-C | Keyboard: Ducky One 3 TKL (Cherry MX-Speed-Silver)Beyerdynamic MMX 300 (2nd Gen) | Acer XV272U | OS: Windows 11 |

Link to comment
Share on other sites

Link to post
Share on other sites

The real time chat capabilities are very impressive. It has a few hiccups now and then, but seeing how quickly these LLMs have improved over the last bit, it's naive to think they can't iron out these issues.

 

I haven't used Google in years at this point and don't miss it one bit. Bing with the ChatGPT integration is so much better. It takes some getting used to, but just using voice input to get quick information without having to skim through 5 different sites is something else. Getting rid of the turn based interaction is a huge step forward.

 

Knowing what GPT-4o can do really makes the Amazon Echo dot I have at home and pretty much all other voice assistants look like children's toys.

 

By default, it's just a little too excitited for my taste.

 

I also find it funny that a Microsoft-owned company is doing their keynote on Apple devices.

If someone did not use reason to reach their conclusion in the first place, you cannot use reason to convince them otherwise.

Link to comment
Share on other sites

Link to post
Share on other sites

From my experience so far, the best part of GPT-4o is its translation speed. Being able to translate massive documents in seconds is incredible. It's twice as fast as turbo. I often translate 100's sometimes 1000's of pages of legal documents for my job and this model saves me so much time.

GPU: XFX RX 7900 XTX

CPU: Ryzen 7 7800X3D

Link to comment
Share on other sites

Link to post
Share on other sites

2 hours ago, Orangeator said:

From my experience so far, the best part of GPT-4o is its translation speed. Being able to translate massive documents in seconds is incredible. It's twice as fast as turbo. I often translate 100's sometimes 1000's of pages of legal documents for my job and this model saves me so much time.

I've also found ChatGPT to be a lot better at translating than for example Google Translate, at least for Japanese which heavily relies on context. Google Translate handles context very poorly, while ChatGPT does it really well.

Link to comment
Share on other sites

Link to post
Share on other sites

Is there any Information about if GPT-4o will be integrated into bing search, Copilot or Windows in general? So far it seems like it's only going to be available through the ChatGPT apps.

If someone did not use reason to reach their conclusion in the first place, you cannot use reason to convince them otherwise.

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×