Launching today
Ertas
Build on-device AI for your app. No ML expertise required.
3 followers
Build on-device AI for your app. No ML expertise required.
3 followers
Build a custom AI model and deploy on-device in your iOS, Android, or desktop app. Zero per-inference cost, works offline. No ML expertise required. Launch Edition plans opens on May 22 and goes away 72 hours after launch. Use promo code LAUNCH50 for 50% off standard price forever.
Hi everyone, Edward here, founder at Ertas.
Two years ago I shipped my first AI-powered mobile app. Eight thousand people used it. The OpenAI bill that month was $620. I shut the AI feature off the next morning.
That moment is why we built Ertas.
Ertas is a visual canvas for fine-tuning open AI models on cloud GPUs (Gemma 4, Qwen 3.5, Llama 3.2, Phi-4 mini, etc). You upload / craft your data, pick a base model, train, evaluate against your real prompts, and export GGUF. Then you ship the model on-device in your iOS, Android, or desktop app. Zero per-token cost forever. Works offline. User data stays on the phone.
Real numbers from our design partners: 94% domain accuracy on B2B SaaS task categorisation (vs 71% with prompt-engineered GPT-4), 87% support chatbot auto-resolution (vs 34% with RAG + frontier API).
Today is launch day. Free tier opens with 5 credits a day, no card required. Launch Edition offer opens now. 50% off forever for Builder, Pro, and Business plans. The window is 72 hours, then it closes.
I will be in the comments here, on LinkedIn, and on X for the rest of the day. Drop your hardest technical questions, your weirdest fine-tuning use cases, and the thing in your app that is currently a per-token nightmare.
Thanks for landing here.
www.ertas.ai