Build models that you trust.
Build a custom model for your app. Deploy on-device. Zero per-inference cost.
Build your AI model. Visually.
How it works
From your data to on-device AI in three steps.
Upload your dataset
Upload your JSONL training dataset through our web interface. Don't have a dataset? Import one directly from Hugging Face using a URL.
Fine-tune in Studio
Fine-tune multiple models simultaneously on fast cloud GPUs — no expensive hardware needed. We save the knowledge of your fine-tuned models, enabling iterative testing across different use cases.
Ship in Your App
Export your fine-tuned model as GGUF and ship it in your iOS, Android, or desktop app. Run inference on-device with llama.cpp. Zero per-token cost, works offline.
Roadmap
Our path to making AI fine-tuning accessible to everyone.
Studio
Upload your data and fine-tune pre-trained models on fast cloud GPUs — no expensive hardware required. Your datasets are stored for convenience so you can iterate quickly. Smart data synthesis suggestions coming soon to enrich your training data.
Why fine-tune your own model?
Custom models give you capabilities that prompting a generic LLM can't match.
Models that understand your domain
Fine-tuned models outperform generic LLMs at your specific tasks — better than any prompt engineering.
Costs that don't scale with your users
Cloud APIs charge per token, and your bill grows with every user. On-device inference costs nothing at runtime, whether you have 1,000 users or 100,000.
Ship AI features without ML expertise
Ertas handles the training pipeline. You focus on your product, not infrastructure.
On-device: offline, instant, private
No network round-trip, no loading spinners, no data leaving the device. Your AI features work everywhere, even offline.
Built for app builders who ship AI-powered products: indie devs, startups, agencies, and solo builders.
"I fine-tuned a model on our product docs in under an hour. Now our support bot actually understands our domain instead of hallucinating."
Jamie K.
Indie Developer
"Replaced our $400/mo API bill with a fine-tuned model running locally. Better results, predictable costs. Exactly what we needed."
Maria R.
Startup Founder
Built for how you work
Pre-subscribe to start building on-device AI in the private beta today — free until launch. Lock in early bird pricing for life.
Free
Try it out — 5 credits/day
- 30 credits/mo (5/day)
- 1 project
- 5 GB model storage
- Up to 7B models
- Manual testing sandbox
Builder
Early bird price locked for life
For indie devs and solo builders
- 100 credits/mo
- 3 projects
- 50 GB model storage
- Up to 14B models
- Dataset synthesis (5 runs/mo)
- Bulk eval (100 test cases/run)
- 3 auto-evals/mo
- Credit rollover (1 month)
Includes free private beta access with daily-refreshed credits
Billing starts at launch. Cancel anytime for a full refund.
Pro
Early bird price locked for life
For power users and teams managing client work
- 400 credits/mo
- 10 client-labeled projects
- 200 GB storage
- 5 team seats
- 3 concurrent fine-tuning jobs
- Dataset synthesis (20 runs/mo)
- Bulk eval (500 cases + objectives)
- Higher tier GPUs
Includes free private beta access with daily-refreshed credits
Billing starts at launch. Cancel anytime for a full refund.
Business
Early bird price locked for life
For scaled teams and multi-client agencies
- 1,000 credits/mo
- 30 client-labeled projects
- 500 GB storage
- 15 team seats
- 8 concurrent fine-tuning jobs
- Unlimited dataset synthesis & eval
- Bulk eval (2,000 cases + regression)
- White-label API access
Includes free private beta access with daily-refreshed credits
Billing starts at launch. Cancel anytime for a full refund.
Need compliance, on-prem, or frontier hardware? Talk to us about Enterprise
Frequently asked questions
Everything you need to know about Ertas AI.
Ship AI that runs on your users' devices.
Build your model on cloud GPUs, deploy on-device, pay nothing per inference.