Today is a huge day for developers. 🤯
- ChatGPT API released (10x cheaper)
- Whisper available in the API
- Overhauled data usage policy
- Focus on stability
And more!
Here’s a quick thread on everything we shipped today
@OpenAI
🧵
Excited to share I’ve joined
@Google
to lead product for AI Studio and support the Gemini API.
Lots of hard work ahead, but we are going to make Google the best home for developers building with AI.
I’m not going to settle for anything less.
Yesterday was my last day
@OpenAI
.
I spent the last year and a half putting my heart and soul into supporting developers every single day.
I’m going to miss all the amazing folks I got to work with, y’all are doing such important work, keep shipping ♥️
Stay tuned…
ChatGPT Plugins are here 🤯
Today we are launching:
- Browsing in ChatGPT
- Code execution in ChatGPT
- 3rd party plugins
And a whole new developer ecosystem! Here's what you need to know 🧵
Today, we are rolling out three experimental models:
- A new smaller variant, Gemini 1.5 Flash-8B
- A stronger Gemini 1.5 Pro model (better on coding & complex prompts)
- A significantly improved Gemini 1.5 Flash model
Try them on , details in 🧵
A note to
@OpenAI
developers 🫶:
I wanted to express my appreciation for all the warm, thoughtful, and supportive messages I got and I’ve seen posted across the community.
Despite a moment of uncertainty, our commitment to developers remained steadfast.
In the meantime,
Beyond excited to share that today is my first day
@OpenAI
where I’ll be their first Developer Advocate and helping lead/build Dev Rel! 🥳
I’ll be supporting the developer community using/building with ChatGPT, GPT-3, DALLE, the API, and more! 🚀
PSA: You can get started building with the Gemini API in less than 5 minutes, no credit card required, with just an API key, and get >1500 requests per day for free! 🏎️
NotebookLM updates (since people are loving it):
- You can now share Audio Overview via a public URL, accessible to everyone by default
- Support added for YouTube videos and audio files as new source materials
Enjoy : )
Two new production Gemini models, >2x higher rate limits, >50% price drop on Gemini 1.5 Pro, filters switched to opt-in, updated Flash 8B experimental model, and more.
It’s a good day to be a developer : )
Starting today, you can set custom instructions in ChatGPT that will persist from conversation to conversation. 👀 📌
You can enable custom instructions in the beta panel from the settings.
I have so much love and respect for my colleagues at
@OpenAI
.
Truly the most incredible group of people on the planet. This has been an utterly devastating last 3 days.
Mira deserves to take a break, she’s been sprinting at OAI for 5+ years, and also to get a chance to build 0-1 again.
Everything doesn’t need to be a controversy.
Say hello to Gemini 1.5 Flash-8B ⚡️, now available for production usage with:
- 50% lower price (vs 1.5 Flash)
- 2x higher rate limits (vs 1.5 Flash)
- lower latency on small prompts (vs 1.5 Flash)
I have been putting every ounce of my life over the last 5 months into making the Gemini API and Google AI Studio the best products for developers building with AI.
Gemini 1.5 Flash-8B is wild:
- 1,000,000,000 cached tokens for $10
- 1 million token context window
- Native multimodal support (image, audio, video)
- Code Execution, Structured outputs, etc
Truly in a class of its own
Good news for
@GoogleAI
developers:
- Gemini 1.5 Flash price is now ~70% lower ($0.075 / 1M)
- Gemini 1.5 Flash tuning available to all
- Added support for 100+ new languages in the API
- AI Studio is available to all workspace customers
- Much more : )
We’re thrilled to release a major new version of our TypeScript / Node SDK for the OpenAI API! 🔥
Version 4 offers a huge set of improvements – some of the highlights include:
- Streaming responses for chat & completions
- Carefully crafted TypeScript types
- Support for ESM,
Google was first to ship:
- 1 M context window (and 2 M)
- A SOTA multi-modal LLM
- Context Caching
- A high quality small model for developers (Flash)
- Both SOTA proprietary and open model lineup
- More to come
So yeah, definitely no innovation happening here…..
We just shipped a series of changes which have significantly improved the Gemini 1.5 Flash latency (>3x reduction) and output tokens per second (>2x more)⚡️🚢
We just increased the max PDF page upload size to 1,000 pages or 2GB (up from 300 pages) in Google AI Studio and the Gemini API. 🗒️
We use both text understanding and the native multi-modal capabilities of Gemini to process these documents (1 image per page).
New
@Google
developer launch today:
- Gemini 1.5 Pro is now available in 180+ countries via the Gemini API in public preview
- Supports audio (speech) understanding capability, and a new File API to make it easy to handle files
- New embedding model!
In the next 10 years we are going to have:
- super human AI
- full self driving everywhere in the world
- humans on mars
- internet everywhere on earth
- supersonic commercial jets
- cures for major diseases
Keep building, there’s still more to do 🚀
Personal news: I have joined
@Wharton
as a Senior Fellow working with
@emollick
(outside my role at Google)
Ethan is genuinely pushing the frontier on understanding and shaping AI’s impact on education. Excited to work with him!
People are still underestimating the value of Gemini 1.5 Flash.
For $0.35, you can get 1 million tokens and start building natively multi-modal projects.
The cost + latency + context window size + intelligence of Flash is going to create so many new startups.
Incredible news for
@OpenAI
devs:
- new GPT-4 and 3.5 Turbo models
- function calling in the API (plugins)
- 16k context 3.5 Turbo model (available to everyone today)
- 75% price reduction on V2 embeddings models
And more 🤯🧵
Good news, the
@OpenAI
fine-tuning UI now supports end to end job creation all in the UI, no code required to kick off a job! 🤯
Democratizing access to fine-tuning the worlds most advanced models is a huge win.
Congrats to
@slessans
on the ship! 🎉
The
@OpenAI
fine-tuning UI is here! 🔥
You can now see your fine-tunes directly and will be able to create them though the UI in the months to come!
We also bumped the concurrent training limit from 1 to 3 so you can fine-tune more models!
We just shipped a bunch of new Gemini API parameters 🚢:
- logprobs
- candidateCount
- seed
- presencePenalty
- frequencyPenalty
- model version in response
More coming soon, along with a slew of doc updates to go into more details!
Siri becoming 10 - 100x more useful is genuinely going to change so many people’s lives.
Really excited for Apple to get to deliver this experience to their customers (including me).
Exciting News from Chatbot Arena!
@GoogleDeepMind
's new Gemini 1.5 Pro (Experimental 0801) has been tested in Arena for the past week, gathering over 12K community votes.
For the first time, Google Gemini has claimed the
#1
spot, surpassing GPT-4o/Claude-3.5 with an impressive
On the way back from the office today, almost got into a car accident when I saw this, I had to pull off and take a picture. 🤯
Am I being pranked?
#JuliaLang
Today, we are making an experimental version (0801) of Gemini 1.5 Pro available for early testing and feedback in Google AI Studio and the Gemini API. Try it out and let us know what you think!
Huge news, ChatGPT for Enterprise is now available and features:
- Unlimited access to GPT-4 (no usage caps) 🤯
- Higher-speed performance for GPT-4 (up to 2x faster)
- Unlimited access to advanced data analysis (code interpreter)
- 32k token context windows for 4x longer
Hot take 🔥: ChatGPT and LLM’s have created more value in 6 months than Crypto has since inception.
Not a dig on crypto, but a case study in how to solve the real problems users have in an accessible way.
We just shipped a new variant of Structured Outputs in the Gemini API called Enum Mode, which allows you to easily constrain the model to pick between pre-defined options 🚢
Want to see some ChatGPT Plugin examples? 🔌👀
I created a GitHub repo for the community to submit examples of fully working ChatGPT Plugins in multiple programming languages:
Good news for OpenAI devs 📣
You can now name your API keys! This has been a highly requested feature for a while now, glad it has landed. S/o to our engineering team for making it happen. 👏
Great news for
@Google
developers:
Context caching for the Gemini API is here, supports both 1.5 Flash and 1.5 Pro, is 2x cheaper than we previously announced, and is available to everyone right now. 🤯
Exciting news for
@OpenAI
devs: we are close to a 1.0 release of the OpenAI Python SDK 🎊. You can test the beta version of 1.0 today, we would love to get your early feedback!
We just shipped a new set of evals measuring long context reasoning performance which are challenging for frontier models and go beyond needle-in-the-haystack measurement.
1.5 Pro is 12.5% (1.5 Flash is 6.8%) better than the next closest models. 🧮
Huge news for
@OpenAIDevs
: API key based usage is here 🥳
To get started, head to the API key page and generate a tracking token for each key which will enable per key tracking in the usage dashboard for all new requests!
Despite what you see here on Twitter and all the hype, GPT-5 is not being trained right now, nor will it be for some time.
GPT-4 took 6 months post training to ensure a safe and aligned model. This is likely to increase for future models.
TLDR; focus on GPT-4!
You could make a lot of money 💰 right now with a Generative AI / Large Language Model consultancy integrating
@OpenAI
into products and services.
Build a diverse portfolio of examples and there will be unlimited demand.
People still fail to realize how wild long context is. In practice, 2 million tokens looks like:
- 100,000 lines of code
- All the text messages you have sent in the last 10 years
- 16 average length English novels
- Transcripts of over 400 podcasts
🤯🤯🤯
ChatGPT for teams is now available! It comes with:
- Higher message caps
- GPT-4 32k
- No training on your conversations
- Workspace and team management
- Early access to new features
- GPTs shared in your workspace
All for $30 per month 🥳
Great news for ChatGPT users, starting today, we are releasing new ways to manage your data 💽.
You can now turn off the Chat history feature which will make sure the conversation is not used to train our models nor saved in the UI.
More info in 🧵
We just shipped a new native prompt gallery in Google AI Studio ✨
Test out long context, native multi-modal (image, video and audio), structured outputs, and more!
Hot take 🔥: Many believe prompt engineering is a skill one must learn to be competitive in the future.
The reality is that prompting AI systems is no different than being an effective communicator with other humans.
The same principles apply in both cases. This makes me
Awesome new
@OpenAI
developer resources just dropped 👀:
- GPT guide
- GPT best practices (prompt engineering)
- Updated introduction
And more! If you use large language models, the GPT best practices is a must read! More info in 🧵
New: excited to publicly share the
@OpenAI
Forum, a place to discuss, learn, and shape AI.
The forum features online and in-person events along with paid activities that directly impact OpenAI models.
Details in 🧵, along with how to join. (1/n)