Pricing
Workspace plans
Prices are in US dollars. Each billing period includes usage credit. Both plans charge the same rates for managed inference and training.
| Pay-as-you-go | Scale | |
|---|---|---|
| Monthly subscription | $0 | $249 |
| Included monthly usage credit | $10 | $50 |
| Included storage | 1 GB | 5 GB |
| Storage above allowance, per GB-month | $0.10 | $0.10 |
| Default inference requests per second | 80 | 160 |
| Default input tokens per second | 100,000 | 250,000 |
Pay-as-you-go includes community Slack support. Scale adds priority production support, private Slack support, and access to the SOC 2 Type II report. Zero Data Retention is enabled by default on every plan for synchronous direct and custom-model inference; capture defaults off.
Enterprise offers volume discounts or machine-time pricing, managed or customer-cloud deployments, SSO, and dedicated 24/7 support with a custom SLA. Contact sales for Enterprise pricing.
Managed inference
| Model | Input per million tokens | Output per million tokens |
|---|---|---|
| Mo | $0.04 | Free |
| Mo Lite Coming soon | $0.03 | Free |
Mo Lite is coming soon and is not available for inference yet. Requests using mo-lite, mo-lite-latest, or mo-lite-1.0 (including their modaic/ aliases) return HTTP 400 with code model_not_supported. Use mo-latest or mo-2.0 for inference.
Managed training
| Training method | Context length | Price per million training tokens |
|---|---|---|
| LoRA | Up to 64k | $5.11 per epoch |
| Head-only | Up to 64k | $0.75, counted once |
| Head-only | Up to 256k | $1.50, counted once |
For example, a LoRA fine-tune on 1 million training tokens for 3 full epochs costs $15.33 before credits. Estimate a run with the pricing calculator.
Storage
Each workspace has one continuous storage allowance shared across its models. Storage above that allowance costs $0.10 per GB-month, prorated by the number of bytes stored and how long they are kept. A GB is one billion bytes. Storage charges appear as a separate monthly invoice line. Enterprise percentage discounts follow your contract.
In Modaic Platform, open Billing & Usage → Usage to see checkpoints, model weights, training data, files and datasets, images, examples, annotations, and predictions. The Model dropdown filters both token usage and storage. Select All to include shared workspace files. A custom-model selection shows storage directly attributable to that model. Shared base models show zero customer storage.
Saved examples, annotations, and predictions are measured by the UTF-8 bytes of their current retained text and JSON content. Deleted records and superseded versions are excluded from that measurement. Usage totals update as storage is inventoried.
Paused and completed fine-tuning jobs retain checkpoints indefinitely. Retained checkpoints contribute to storage usage. Cancelling an active or paused job deletes its saved optimizer checkpoints. Cancelling a completed job leaves its checkpoints unchanged and they continue to count toward storage usage. For completed jobs, delete individual checkpoints through the API. See the fine-tuning job reference.
Usage credits
Included credit expires at the end of each billing period. You can buy $5–$500 in additional credit per top-up or enable automatic recharge. Purchased credits expire after 12 months. Requests pause when your balance reaches zero and resume when credit is available.