Build on-device AI in your app. Fine-tune open models on cloud GPUs, export GGUF, ship to iOS, Android, and desktop. Zero per-inference cost. Works offline. User data stays on the device.
We're building the fine-tuning + deployment runtime for the 2026 reality: agentic apps burn 5 to 30x more tokens than chatbots, and the API bill cliff hits between 500 and 5,000 users. On-device is the answer most builders haven't shipped yet because the last mile (GGUF to a working app) is 20 to 40 hours of llama.cpp plumbing. We close that gap.
Open base models (Gemma, Qwen, Llama, Phi, gpt-oss), memory-efficient LoRA training kernels, llama.cpp runtime, GGUF export. Nothing forked, nothing hidden, your model runs anywhere llama.cpp runs, including without us.
We're builders at heart. We've shipped products that leaned on AI, hit the walls every builder hits (API costs, deployment, accessibility, time), and worked around them the long way. Ertas is the runtime we wish we'd had then.
We're building this especially for builders. The mobile app developer. The indie hacker. The agency. The operator. The builders who shouldn't have to spend 20 to 40 hours wiring llama.cpp into a project to ship one fine-tuned model into their app.
Humans behind Ertas: Edward Yang (CEO) · Franco Jimenez (CTO) · Ani Salunke (CPO)
Friday 2026-05-22 at 00:01 PST. Free tier opens the same day. No card required.
The first 72 hours are an Early Bird window. Pricing during the window locks for the lifetime of your subscription, even as the product evolves:
| Tier | Early Bird (May 22 to 25) | Standard (after) |
|---|---|---|
| Builder | A$14.50/mo | A$34.50/mo |
| Pro | A$69.50/mo | A$149/mo |
| Business | A$169/mo | A$349/mo |
Want to lock the price before the public launch? Pre-subscription is open now at https://www.ertas.ai. Free beta access, daily-refreshed credits, full billing starts on launch day.