Summary
A developer has successfully run a 28.9-million-parameter language model, trained on the TinyStories dataset, on a $10 ESP32-S3 microcontroller, generating text at nearly 10 tokens per second offline. This was achieved by optimizing memory usage, including storing most parameters in flash storage, making it a notable advancement for running LLMs on limited hardware.
Turn this story into an image post, carousel, video, audio reel, blog, and more — all from HyperFast News.
Pick a story, apply your brand template, and publish polished graphics to Facebook, Instagram, LinkedIn, and more — without opening a design tool.
Create Image PostDescribe the story, headline, colors, and style you want.
Try Prompt to PostCombine branded news graphics with narration — from your audio library, a live recording, or AI text-to-speech — and share as engaging short-form content.
Create Audio PostUpload your footage, apply a video template with baked-in headlines, add optional audio, and schedule posts across every connected social channel.
Create Video PostCombine several news items into a single swipeable carousel post — with a shared title, unified caption, and per-slide template content for maximum engagement.
Create CarouselUpload your media, craft the perfect caption with AI assistance, preview on every platform, and schedule posts for the exact moment your audience is online.
Schedule a PostTransform breaking news into long-form, SEO-optimized blog posts with AI writing, a rich text editor, featured images, and one-click social image post creation.
Create Blog PostTurn any news article into a professional anchor video with AI-generated presenters, natural voice narration, and lip-sync animation — no studio, no crew, no wait.
Learn moreAbout the public HyperFast News news feed powered by NewsEngine.
