Mystic AI

Today, we’re thrilled to announce the launch of our groundbreaking solution to the cold start problem for AI models in the cloud. Our engineering team has built a custom Docker registry and a containerd adapter that revolutionise the loading process. Written in Rust, our new Turbo Registry leverages the full networking and CPU capabilities of machines to achieve record-breaking container load times. 5GB, 10GB, and 20GB docker images now load in just 10.23, 14.75, and 23.72 seconds, compared to the traditional 82.21, 147, and 270.47 seconds. As AI model sizes and parameters grow, Turbo Registry ensures the efficiency and speed needed to keep pace. Experience the future of container loading with our cutting-edge solution. https://lnkd.in/dGtH94eD

💥 Super excited to announce our new official partnership with Black Forest Labs, creators of Flux; The new state-of-the-art image generation model, also used in X’s Grok to generate images. You can now access the Flux models commercially on Mystic AI. We’re one of only three official API providers! 🎉 And the only provider that charges for inference time, so better pricing for varying quality/speed. The models included are: - Flux Pro (Highest quality but slower) - Flux Dev (Best quality/performance)  - Flux Schnell (Good quality & fast, specially with low steps) 🌲 Black Forest Labs & Mystic AI will be releasing new stuff shortly, follow to stay up to date! Run here: https://lnkd.in/eGbJhjBR Read more about the models here: https://lnkd.in/eCWTgucP (All images below are generated with Flux) #AI #MachineLearning #ImageGeneration #BlackForestLabs #Flux

👀 Did you know you can run your own ComfyUI workflows as an API on Mystic? This is a generated video of my dream home, made with ComfyUI and Mystic using Flux, image upsampler and the video generated with SDV. All open-weights! ComfyUI is used by thousands of developers today to prototype and launch workflows of text2img, text2video, image upsampling, inpainting, etc. If you are looking to build your own AI product in this space, you’ve probably heard of it. The problem comes at scale. Say you’ve build an awesome workflow and you want 100s of users to access it too, now you have to figure out how to scale this ComfyUI workflow. 👎 Some of the challenges, - Requires expensive GPUs for fast inference. - Docker image is large (>60 GB), making cold starts extremely long of several minutes. ✅ We optimised this at Mystic AI with, - Running your workflow on spot GPUs (Up to 10x cheaper than on-demand GPUs) - Our Rust-based Turbo Registry, lowering cold-start by 14x, from minutes to seconds! Learn how to deploy your ComfyUI-as-api in this tutorial here: https://lnkd.in/eXCAtUcH #comfyui #serverless #machinelearning #ML #AI #GPUs #StableDiffusion #Flux

🔥 The Mystic Turbo Registry is now live on Product Hunt! We would love for you to continue supporting our endeavours by sharing your opinions and valuable feedback and help us spread the word. Our Turbo Registry, loads models upto 15x faster - cutting down cold start times by up to 90% Check us out on Product Hunt now! https://lnkd.in/ezr-dj-P Paul H. Oscar Rovira James Dowling yvan buggy Matthew Cherry Adam Leith P Ross Gray Anirudh Raghuram

Every time I look at this video I’m amazed at the work Mystic AI team keeps doing to push the boundaries of ML infrastructure. I'm talking about our new Turbo Registry. An optimized Rust-based Docker registry. This is a 13GB Docker Image (i.e., SDXL and all libraries and packages required to run) loaded into a new GPU in 12s! 💥 (Instead of the standard 54s 💩). The best news is that it scales nicely, with improvements of up to 15x for larger Docker images. This means that if you deploy your ML model with Mystic, going from 0 GPU to 1 would take a few seconds instead of close to 1 minute. That is 1 minute less your customers have to wait until they get a response back from your AI product. #ML #AI #MachineLearning #MLOPs #LLMOPs

We are excited to soon roll out the fastest Docker cold starts in the world to our customers! By utilising all available server resources, we can download containers at record speed, making them ready to start within seconds without any decompression required. Check out this video showing a 12GB Stable Diffusion model container being downloaded in under 10 seconds—a process that typically takes several minutes. We’re confident we can achieve even faster speeds in the near future. Stay tuned for more updates, and join the waitlist here: https://lnkd.in/et34aspp.

Been a month since London Tech Week 2024, here's a quick look back... Oscar Rovira

Been a month since London Tech Week 2024, here's a quick look back... Oscar Rovira

Over the past few weeks, our engineering team has been working really hard to build our proprietary Turbo Registry. Here's a few snapshots of Paul H. demoing it live at AI Tinkerers London last night. Cold starts are one of the main problems when running AI models in the cloud. Our new Turbo Registry and containerd plugin enable the full networking and CPU capabilities of a machine to be utilised to load containers in record-breaking times (loads a 10GB Docker image in less than 15 secs). Join our waitlist to be among the first to access this feature: https://lnkd.in/d8UMwbtv

Mystic AI has achieved record-breaking speed when loading AI models with docker. At speeds x8 faster than every cloud provider, we can load AI based Docker containers in under 15 seconds—this typically takes several minutes. The model loading (cold-start) problem is seen across the industry, but there's been little progress. When models load slowly, it ramps up server bills and creates a terrible user experience for your products. We're rolling out this new acceleration to our users over the coming weeks; you can sign up for the waitlist to be among the first to access it here: https://lnkd.in/et34aspp. For those interested in how: we created a new type of docker registry written in Rust, then rewrote how Kubernetes and containerd load images. This will be open-sourced later this year. #ai #rust #kubernetes #k8s #mystic #docker #llm #genai