Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

We took our most efficient model and made an open-source iOS app📱but why? As phones get faster, more AI will happen on device. With OLMoE, researchers, developers, and users can get a feel for this future: fully private LLMs, available anytime. Learn more from Luca Soldaini 🎀👇

Ai2

85,732 subscribers

234,956 views • 1 year ago •via X (Twitter)

Science & Technology News & Politics Education

Anya Rossi• Live Now

Private livecam show

11 Comments

Ai21 year ago

If you have an iPhone 15 Pro or later, or iPad with Apple Silicon, download OLMoE now: If you’re a researcher or iOS dev, head over to GitHub: Our code works with any GGUF model. Use it as the starting point for your new app!

Ai21 year ago

Finally… bonus release 🎉 We refreshed our models using OLMo 2 mid-training and Tülu 3 post-training recipes! OLMoE-0125 bests OLMoE-0924 by 8.4 points on average, and OLMoE-0125-Instruct is 12.6 points better. All artifacts available on Hugging Face 🤗

Ai21 year ago

For more behind-the-scenes, head over to our blog: We can't wait to hear how you are using OLMoE! Let us know on Discord 👀

Sabine VanderLinden1 year ago

A year ago, I saw some clients struggling with fragmented data and slow decision-making, holding their business back. 🌍 But what if there was a way to harmonize everything—processes, data, and AI—to move smarter and faster? That’s why I’m interested in @SAP’s upcoming event.

Pasquale Minervini is on 🦋1 year ago

@soldni Things are getting out of hand 😅

Cavit Erginsoy1 year ago

@soldni You guys set the bar really high about how to do a model release👏 The app in swift being apache 2 is even more impressive. One gripe I have is prioritising llama.cpp with no mlx support, but no doubt it's in the pipeline and now that it's out I'm sure it'll follow in no time

merve1 year ago

@soldni brb buying an iPhone 16 Pro

Luca Soldaini 🎀1 year ago

chat what's happening with those brows tho👀

Josmendez- Josh1 year ago

@soldni Please make a version for Android although it is not easy the diversity of hardware is also a positive thing, there are devices with more power of execution on the iphone

λux1 year ago

@soldni running olmoe on my ipad pro 🏃🏻‍♂️

Anshu1 year ago

@soldni Finally a good reason to upgrade from my old iPhone 12 to a new one !!

Related Videos

The future of AI is open-source. And ollama is the easiest way to build AI applications with open-source LLMs. Here's how to build a free, private RAG app using open-source tools. We'll use: - Ollama for LLMs and embedding models - PostgreSQL for data storage and retrieval - pgai Vectorizer for embedding creation and sync (I use Nomic for embeddings and tinnyllama as my LLM but you can substitute them for any models on Ollama)

The future of AI is open-source. And ollama is the easiest way to build AI applications with open-source LLMs. Here's how to build a free, private RAG app using open-source tools. We'll use: - Ollama for LLMs and embedding models - PostgreSQL for data storage and retrieval - pgai Vectorizer for embedding creation and sync (I use Nomic for embeddings and tinnyllama as my LLM but you can substitute them for any models on Ollama)

Avthar

34,261 views • 1 year ago

How can we use small LLMs to shift more AI workloads onto our laptops and phones? In our paper and open-source code, we pair on-device LLMs (ollama) with frontier LLMs in the cloud (OpenAI, Together), to solve token-intensive workloads on your 💻 at 17.5% of the cloud cost while maintaining 97.9% of the accuracy. See Gru and the Minions in action below, 🔉on please (h/t )!

How can we use small LLMs to shift more AI workloads onto our laptops and phones? In our paper and open-source code, we pair on-device LLMs (ollama) with frontier LLMs in the cloud (OpenAI, Together), to solve token-intensive workloads on your 💻 at 17.5% of the cloud cost while maintaining 97.9% of the accuracy. See Gru and the Minions in action below, 🔉on please (h/t )!

Dan Biderman

193,452 views • 1 year ago

Hugging Face CEO Clem Delangue predicts that a growing share of AI inference will run locally, on-device. It's faster. Cheaper. Private by default. And as open source AI moves from apps to machines, it could trigger “some sort of ChatGPT moment for robotics.” The next AI breakthrough will happen where intelligence becomes embodied, cheap, private, and open.

Hugging Face CEO Clem Delangue predicts that a growing share of AI inference will run locally, on-device. It's faster. Cheaper. Private by default. And as open source AI moves from apps to machines, it could trigger “some sort of ChatGPT moment for robotics.” The next AI breakthrough will happen where intelligence becomes embodied, cheap, private, and open.

vitrupo

37,764 views • 1 year ago

The future of Play is everywhere your users are. 🌍 From app discovery in the Gemini app to agentic catalog management in Play Console, we are giving you the tools to scale faster. We caught up with attendees to get their immediate reactions - watch here! 👇 and learn more:

The future of Play is everywhere your users are. 🌍 From app discovery in the Gemini app to agentic catalog management in Play Console, we are giving you the tools to scale faster. We caught up with attendees to get their immediate reactions - watch here! 👇 and learn more:

Google Play business community

23,528 views • 2 months ago

Every time I get to talk with Nikola Jokić, I learn about and appreciate him as a player and him as a person even more. This is a clip from our long sit-down together, available on the Altitude+ app

Every time I get to talk with Nikola Jokić, I learn about and appreciate him as a player and him as a person even more. This is a clip from our long sit-down together, available on the Altitude+ app

Katy Winge

17,500 views • 5 months ago

The Pyra mobile app is here 🔥🔥🔥 Hold your crypto, earn yield on it, and spend without capital gains with our DeFi-powered neobank. After a successful private beta, we're now live on iOS and Android! First 100 users get free onboarding. Learn more here 🫳

The Pyra mobile app is here 🔥🔥🔥 Hold your crypto, earn yield on it, and spend without capital gains with our DeFi-powered neobank. After a successful private beta, we're now live on iOS and Android! First 100 users get free onboarding. Learn more here 🫳

Pyra

63,393 views • 1 year ago

This is Eliza The fastest, most powerful, most consumer friendly, most open source agent you’ve ever used Coming to iPhone, Android, browser, desktop and as a full operating system on funky devices like this one (LightPhone 3) You can get this from our GitHub right now, and in app stores everywhere very soon By the way, we’re raising and we’re looking for passionate builders who want to make sure the future is private, in our hands, open source and owned by us

This is Eliza The fastest, most powerful, most consumer friendly, most open source agent you’ve ever used Coming to iPhone, Android, browser, desktop and as a full operating system on funky devices like this one (LightPhone 3) You can get this from our GitHub right now, and in app stores everywhere very soon By the way, we’re raising and we’re looking for passionate builders who want to make sure the future is private, in our hands, open source and owned by us

Shaw (spirit/acc)

55,444 views • 3 days ago

While we in India should be cautious with the DeepSeek app, we can totally make use of the open source model namesake, if securely deployed on Indian servers, to leapfrog our own AI progress. कृत्रिम has deployed DeepSeek-R1 671B on H100s - first time anywhere in the world. It’s the most powerful open-source model available and to enable all Indian developers to access it, we will price it at ₹1/million tokens for February.

While we in India should be cautious with the DeepSeek app, we can totally make use of the open source model namesake, if securely deployed on Indian servers, to leapfrog our own AI progress. कृत्रिम has deployed DeepSeek-R1 671B on H100s - first time anywhere in the world. It’s the most powerful open-source model available and to enable all Indian developers to access it, we will price it at ₹1/million tokens for February.

Bhavish Aggarwal

108,609 views • 1 year ago

After a month of open Beta optimising we are now officially live! More than 18.000 checkins and 8000 reviews battletested our mobile app! Available for download on iOS, Android and Seeker | Solana Mobile Let's get into it 🧵

After a month of open Beta optimising we are now officially live! More than 18.000 checkins and 8000 reviews battletested our mobile app! Available for download on iOS, Android and Seeker | Solana Mobile Let's get into it 🧵

Lokal

1,070,596 views • 6 months ago

We plug into your account, find what will scale, and build it with AI. Faster learning. Faster iteration. Better performance. We don’t get paid for more ads. Only for winners. Interested? Book a demo on our website.

We plug into your account, find what will scale, and build it with AI. Faster learning. Faster iteration. Better performance. We don’t get paid for more ads. Only for winners. Interested? Book a demo on our website.

Rahul Bhardwaj

1,193,876 views • 3 months ago

Gen-4 Turbo is now available on the Runway API, allowing developers to easily integrate our most powerful and efficient video model directly into their apps and products. Gen-4 Turbo offers the same price, scale and reliability of Gen-3 Alpha Turbo but with the state-of-the-art video generation capabilities of Gen-4. Learn more and get started at the link below.

Gen-4 Turbo is now available on the Runway API, allowing developers to easily integrate our most powerful and efficient video model directly into their apps and products. Gen-4 Turbo offers the same price, scale and reliability of Gen-3 Alpha Turbo but with the state-of-the-art video generation capabilities of Gen-4. Learn more and get started at the link below.

Runway

37,512 views • 1 year ago

💥 Promises made, Promises kept: Bitcoin has come to the App Store📱 Pay for Maple AI using Bitcoin in the App Store starting today! Get the most private cloud AI using freedom money 🤘 Pioneering a new era! Working with Zaprite (₿/⚡), we made Bitcoin available as a payment option in our app and it passed thorough approval by Apple 🦾 iOS app devs, the bitcoin network is now available to you. Implement it yourself, or talk to Zaprite because they make it dead simple 🫡 Let the sats flow!

💥 Promises made, Promises kept: Bitcoin has come to the App Store📱 Pay for Maple AI using Bitcoin in the App Store starting today! Get the most private cloud AI using freedom money 🤘 Pioneering a new era! Working with Zaprite (₿/⚡), we made Bitcoin available as a payment option in our app and it passed thorough approval by Apple 🦾 iOS app devs, the bitcoin network is now available to you. Implement it yourself, or talk to Zaprite because they make it dead simple 🫡 Let the sats flow!

Maple AI

17,722 views • 1 year ago

Revolutionizing Bladder Cancer Diagnosis with ZayaAI We heard you Zayans, From our last videos we understood you all want a much easier to understand explanation of our AI models. So let's dive into the latest breakthrough in AI-powered healthcare, our Bladder Cancer Detection Model. This isn't just another AI tool, it's a game-changer for pathologists. Precision, Speed and Seamless Integration: Traditional diagnosis of urothelial carcinoma (bladder cancer) is often slow and prone to errors. Our AI model automatically detects low and high-grade tumors, along with lymphovascular invasion, with unmatched accuracy and speed. That means faster results, fewer mistakes and a more efficient workflow for medical professionals. Why This Matters: 1.Faster Diagnoses → AI speeds up the process, allowing pathologists to focus on critical cases. 2.Higher Accuracy → Reduces human error and reliance on costly, time-consuming tests. 3.Seamless Integration → No need for complex system overhauls. ZayaAI fits right into existing workflows. The Future of AI in Healthcare: Embracing our AI models isn’t just about using new technology, it’s an investment in a future where healthcare is more precise, efficient and patient-focused. Pathologists get to make faster, more informed decisions and ultimately, patient outcomes improve. This is how AI is redefining medical diagnostics and can help save lives. DeSci + AI = $ZAI ! 🔬⚕️

Revolutionizing Bladder Cancer Diagnosis with ZayaAI We heard you Zayans, From our last videos we understood you all want a much easier to understand explanation of our AI models. So let's dive into the latest breakthrough in AI-powered healthcare, our Bladder Cancer Detection Model. This isn't just another AI tool, it's a game-changer for pathologists. Precision, Speed and Seamless Integration: Traditional diagnosis of urothelial carcinoma (bladder cancer) is often slow and prone to errors. Our AI model automatically detects low and high-grade tumors, along with lymphovascular invasion, with unmatched accuracy and speed. That means faster results, fewer mistakes and a more efficient workflow for medical professionals. Why This Matters: 1.Faster Diagnoses → AI speeds up the process, allowing pathologists to focus on critical cases. 2.Higher Accuracy → Reduces human error and reliance on costly, time-consuming tests. 3.Seamless Integration → No need for complex system overhauls. ZayaAI fits right into existing workflows. The Future of AI in Healthcare: Embracing our AI models isn’t just about using new technology, it’s an investment in a future where healthcare is more precise, efficient and patient-focused. Pathologists get to make faster, more informed decisions and ultimately, patient outcomes improve. This is how AI is redefining medical diagnostics and can help save lives. DeSci + AI = $ZAI ! 🔬⚕️

ZayaAI

18,015 views • 1 year ago

Most capable generalist robotics models today are closed or at best, open weights. But robotics won’t reach its ChatGPT moment without real openness. That GPT moment was built on years of open tools and datasets such as Python, PyTorch, ImageNet and more, that let researchers inspect, reproduce, and build. Today, we’re introducing MolmoAct 2: a fully open-source action reasoning model for real-world robotics. We rethought and reshaped everything! 🧵👇

Most capable generalist robotics models today are closed or at best, open weights. But robotics won’t reach its ChatGPT moment without real openness. That GPT moment was built on years of open tools and datasets such as Python, PyTorch, ImageNet and more, that let researchers inspect, reproduce, and build. Today, we’re introducing MolmoAct 2: a fully open-source action reasoning model for real-world robotics. We rethought and reshaped everything! 🧵👇

Jiafei Duan

110,330 views • 3 months ago

We previously shared our research on Layer Skip, an end-to-end solution for accelerating LLMs from researchers at Meta FAIR. It achieves this by executing a subset of an LLM’s layers and utilizing subsequent layers for verification and correction. We’re now releasing inference code and fine-tuned checkpoints for this work. Model weights on Hugging Face ➡️ More details ➡️ We hope that releasing this work will open up new areas of experimentation and innovative new research in optimization and interpretability.

We previously shared our research on Layer Skip, an end-to-end solution for accelerating LLMs from researchers at Meta FAIR. It achieves this by executing a subset of an LLM’s layers and utilizing subsequent layers for verification and correction. We’re now releasing inference code and fine-tuned checkpoints for this work. Model weights on Hugging Face ➡️ More details ➡️ We hope that releasing this work will open up new areas of experimentation and innovative new research in optimization and interpretability.

AI at Meta

156,598 views • 1 year ago

LLMs can take gigabytes of memory to store, which limits what can be run on consumer hardware. But quantization can dramatically compress models, making a wider selection of models available to developers. You can often reduce model size by 4x or more while maintaining reasonable performance. In our new short course Quantization Fundamentals taught by Hugging Face's @younesbelkada and Marc Sun, you'll: - Learn how to quantize nearly any open source model - Use int8 and bfloat16 (Brain float 16) data types to load and run LLMs using PyTorch and the Hugging Face Transformers library - Dive into the technical details of linear quantization to map 32-bit floats to 8-bit integers As models get bigger and bigger, quantization becomes more important for making models practical and accessible. Please check out the course here:

LLMs can take gigabytes of memory to store, which limits what can be run on consumer hardware. But quantization can dramatically compress models, making a wider selection of models available to developers. You can often reduce model size by 4x or more while maintaining reasonable performance. In our new short course Quantization Fundamentals taught by Hugging Face's @younesbelkada and Marc Sun, you'll: - Learn how to quantize nearly any open source model - Use int8 and bfloat16 (Brain float 16) data types to load and run LLMs using PyTorch and the Hugging Face Transformers library - Dive into the technical details of linear quantization to map 32-bit floats to 8-bit integers As models get bigger and bigger, quantization becomes more important for making models practical and accessible. Please check out the course here:

Andrew Ng

288,304 views • 2 years ago

If you want to learn AI from the experts, keep reading. 💡 Together with UCL, we made a free AI Research Foundations curriculum – available now on Google Skills. With lessons from a Gemini Lead like Oriol Vinyals, you'll explore how to code better, fine-tune an AI model and more. →

If you want to learn AI from the experts, keep reading. 💡 Together with UCL, we made a free AI Research Foundations curriculum – available now on Google Skills. With lessons from a Gemini Lead like Oriol Vinyals, you'll explore how to code better, fine-tune an AI model and more. →

Google DeepMind

287,147 views • 9 months ago

Matte 1.0 is now available on The complete app video studio, from recording to export. Add device frames, zoom effects, and tap indicators. Customize backgrounds, strokes, and more. Try it free, and get 25% off lifetime purchases this week! This early release includes all the basic features but there's a lot more coming! Please report any bugs or issues and expect frequent updates as I work through them. I have a lot of new features planned for the future but feel free to suggest more!

Matte 1.0 is now available on The complete app video studio, from recording to export. Add device frames, zoom effects, and tap indicators. Customize backgrounds, strokes, and more. Try it free, and get 25% off lifetime purchases this week! This early release includes all the basic features but there's a lot more coming! Please report any bugs or issues and expect frequent updates as I work through them. I have a lot of new features planned for the future but feel free to suggest more!

Jose

23,014 views • 6 months ago

on the left: SAM3 on the right: Manako's base model with our new workflow, it takes us 15 minutes to go from left to right AND THEN we get miners to improve the base model, making it more accurate, faster, and cheaper via a fully automated manifest generating process i don't know about you but i love this game

on the left: SAM3 on the right: Manako's base model with our new workflow, it takes us 15 minutes to go from left to right AND THEN we get miners to improve the base model, making it more accurate, faster, and cheaper via a fully automated manifest generating process i don't know about you but i love this game

Max

16,011 views • 5 months ago

Today, we’re thrilled to announce the open weights for Stable Diffusion 3 Medium, the latest and most advanced text-to-image AI model in our Stable Diffusion 3 series! This new release represents a major milestone in the evolution of generative AI and continues our commitment to democratising this powerful technology. 🎉 Learn more and get started here:

Today, we’re thrilled to announce the open weights for Stable Diffusion 3 Medium, the latest and most advanced text-to-image AI model in our Stable Diffusion 3 series! This new release represents a major milestone in the evolution of generative AI and continues our commitment to democratising this powerful technology. 🎉 Learn more and get started here:

Stability AI

308,256 views • 2 years ago