AI_AGI

@googleaidevs The gemini-2.5-flash-preview-04-17 API ignores the deactivation of thought generation — even with budget set to 0, it still delays responses. We want a gemini-2.5-flash as fast and snappy as gemini-2.0-flash, without thought mode.

We’re rolling out implicit caching in the Gemini API. ✨ The feature passes cache savings back to devs without the need to create an explicit cache, providing the same ~75% token discount. Get details here ↓ goo.gle/3GJBcKC

@JeffDean @Google The gemini-2.5-flash-preview-04-17 API ignores the deactivation of thought generation — even with budget set to 0, it still delays responses. We want a gemini-2.5-flash as fast and snappy as gemini-2.0-flash, without thought mode.

The new Gemini 2.5 Pro model is really fun to play with for coding use cases. Check out these nice examples of things people at @Google have been creating.

@OfficialLoganK The gemini-2.5-flash-preview-04-17 API ignores the deactivation of thought generation — even with budget set to 0, it still delays responses. We want a gemini-2.5-flash as fast and snappy as gemini-2.0-flash, without thought mode.

Google AI Studio is barely scratching the surface today of what it will be by the end of the year. Exciting times!

@rez0__ @OfficialLoganK When will the Gemini 2.5 Pro and Flash models be able to support live streaming through the API?

@OfficialLoganK I’m shocked it’s. Or more. Having it in cursor is amazing

Active users in Google AI Studio and the Gemini API have grown over 200% since the beginning of this year, driven by 2.5 Pro 🤯 This is just the beginning…

@OfficialLoganK When are they going to release the native voice they promised for Gemini, and when will they implement it for Gemini 2.0 Pro?

It’s interesting how few people understand that having a super strong base model will lead to significant gains for reasoning as well.

@_aidan_clark_ I sincerely believe that the GPT-4.5 model marks OpenAI’s first clear and undeniable failure—one that has forced us to confront the wall.

I realize I should hype the models not diss the models but I do think some of these quippy "insightful" 4.5isms are a bit cringe.... once a greentext model always a greentext model.....

@apples_jimmy @_aidan_clark_ I sincerely believe that the GPT-4.5 model marks OpenAI’s first clear and undeniable failure—one that has forced us to confront the wall.

@_aidan_clark_ What do you like about 4.5 / what are the improvements people should be talking about ?

@marouane53 واش قدرتي تجرب هاد النموذج فالدارجة المغربية؟ واش لاحظتي شي تحسن فيها بالمقارنة مع جيميناي وكلود؟

After testing GPT 4.5 a bit more, I've realized I don't really have a use for it. If I need something smarter, I would go with the o1 Pro. But for most tasks, even coding, I prefer Claude. Even when it comes to emotional intelligence, which GPT 4.5 claims to have, it doesn't come close to Claude 3.7 For example, if I ask a question that subtly hints at abuse, Claude picks up on that. GPT 4.5 either doesn't understand the real meaning or avoids answering properly. This makes me think we're reaching the end of scaling large language models the old way. AI companies like OpenAI got very lucky they discovered reasoning as a better approach. Because if you look back two years ago at how they used to talk about GPT 4.5 or Claude Opus 4, it's clear that Sam Altman, Dario Amodei, and other CEOs were mostly just hyping things without really knowing how it would turn out. In my opinion, this is OpenAI's first real failure. They no longer have the big lead they once enjoyed. Other companies have models very close to their o1 Pro now. It'll be interesting to see how OpenAI justifies its huge valuation from here.

@christinahkim We PLUS users prefer unlimited Advanced Voice access and higher video/screensharing limits over the SORA model—making our subscription far more valuable.

s2s for all! excited to get out a 4o-like experience out for free users with 4o-mini

@OpenAI We PLUS users prefer unlimited Advanced Voice access and higher video/screensharing limits over the SORA model—making our subscription far more valuable.

Plus users will continue to have access to Advanced Voice powered by 4o with the existing daily rate limit, which is more than 5x the free limit, as well as access to video and screensharing in Advanced Voice. And Pro users will continue to have unlimited access to Advanced Voice and higher limits for video and screensharing.

Starting today, we’re rolling out a version of Advanced Voice powered by GPT-4o mini to give all ChatGPT free users a chance to preview it daily across platforms. The natural conversation pace and tone are similar to the GPT-4o version while being more cost effective to serve.

@AndrewMayne @kimmonismus Andrew. You have to differentiate between canceling the subscription and uninstalling the ChatGPT app.

@kimmonismus Might be an algorithmic thing in your feed. ChatGPT is back on top. Grok is 3 and DeepSeek has sunk to 26.

Wow, I see so many posts that people have cancelled ChatGPT and only use Grok. Grok-3 really is the surprise success par excellence. Outstanding positive feedback.

@sama ChatGPT team, for GPT-4.5/5, consider this: Arabic dialects! Classical Arabic isn't enough. Real-world Arabic varies dramatically. Mastering dialects = reaching millions more users effectively.

really proud of this work:

@OpenAI ChatGPT team, for GPT-4.5/5, consider this: Arabic dialects! Classical Arabic isn't enough. Real-world Arabic varies dramatically. Mastering dialects = reaching millions more users effectively.

Two updates you'll like— 📁 OpenAI o1 and o3-mini now support both file & image uploads in ChatGPT ⬆️ We raised o3-mini-high limits by 7x for Plus users to up to 50 per day

@TheRealAdamG ChatGPT team, for GPT-4.5/5, consider this: Arabic dialects! Classical Arabic isn't enough. Real-world Arabic varies dramatically. Mastering dialects = reaching millions more users effectively.

The hits keep coming.

@BorisMPower ChatGPT team, for GPT-4.5/5, consider this: Arabic dialects! Classical Arabic isn't enough. Real-world Arabic varies dramatically. Mastering dialects = reaching millions more users effectively.

Focus on building companies of the future!

@madebygoogle Could you remove the two user-selected language limit in Gemini's voice chat? Ideally, add automatic language detection to support all languages without pre-selection.

With Gemini Live on Google #Pixel9, there’s a whole new way to talk with your phone: goo.gle/3Q8mbDu #TeamPixel

@madebygoogle Could you remove the two user-selected language limit in Gemini's voice chat? Ideally, add automatic language detection to support all languages without pre-selection.

Have richer conversations with Gemini Live on #Pixel9:¹ 🖼️Add an image and get info, suggestions or help solving problems 📁Reference, summarize and ask questions about PDFs & Google Docs 📺Discuss a YouTube video Keep the convo flowing with Gemini Live: goo.gle/42HugXs

@_lamaahmad @Miles_Brundage ياريت لو يتم تحسين شات جي بي تي في اللهجات العربية، لأنه فعلا ضعيف فيها

@Miles_Brundage It’d be worth it even if I wasn’t

I'm announcing tariffs on all emails entering my inbox

@TheRealAdamG Deepseek efect

@sama more Hype

build monuments in the desert