How Credits Work
Everything you need to know about OTAI training credits — what they are, what consumes them, and how to buy more.
What Are Credits?
Credits are the universal currency for all AI model training operations in OTAI. Every time you train a custom model, credits are consumed. The amount depends on the training mode and the complexity of your attribute set.
Credits are tier-independent — you purchase them on top of your subscription. Your monthly subscription already includes a recurring credit allowance; additional credits can be bought at any time.
Credit Allocation by Tier
Every subscription tier receives credits on sign-up and monthly thereafter:
| Tier | Initial Credits | Monthly Credits |
|---|---|---|
| Community (Free) | 6 | 2 |
| Pro | 100 | 50 |
| Enterprise | 500 | 250 |
Credits do not expire. They accumulate across months — unused credits from one month roll over automatically.
Buying Additional Credits
When your included credits run low, you can purchase more directly in OTAI Studio:
| Package | Price | Credits | Effective Price |
|---|---|---|---|
| 10 Credits | €100 | 10 | €10 / credit |
| 50 Credits | €450 | 50 | €9 / credit (10% off) |
| 100 Credits | €800 | 100 | €8 / credit (20% off) |
What Consumes Credits?
Every training run consumes credits based on two factors:
- Training mode — Basic, Business, or Maximum (quality level)
- Delivery option — Standard or Fast (speed / hardware priority)
Credit Cost per Training Run
| Training Mode | Standard | Fast |
|---|---|---|
| Basic | 1 credit | 2 credits |
| Business | 3 credits | 5 credits |
| Maximum | 12 credits | 18 credits |
Fast uses priority scheduling on NVIDIA H100 hardware for the shortest turnaround. Standard uses NVIDIA L4 hardware with background scheduling at half the credit cost.
Complexity Scaling
The base credit cost assumes a standard attribute set (6–20 attributes). Larger attribute sets scale the credit cost:
| Attributes | Multiplier | Example: Business Standard |
|---|---|---|
| 1–5 | 0.5× | 2 credits |
| 6–20 | 1× (base) | 3 credits |
| 21–50 | 2× | 6 credits |
| 51–200 | 4× | 12 credits |
Training Mode Reference
| Mode | Positioning | Availability | Delivery Time (Fast / Std) |
|---|---|---|---|
| Basic | First working model — evaluation and simple setups | Community, Pro, Enterprise | ~1h / same day |
| Business | Production-ready default — best balance of quality, speed, cost | Pro, Enterprise | ~2h / same day |
| Maximum | Strongest training setup — complex taxonomies, highest accuracy | Enterprise only | ~4h / within 24h |
When Do You Need to Train (and Spend Credits)?
Training is needed when:
- You are setting up a new attribute model (queue routing, priority, category, assigned employee, etc.)
- Your labels change — queues added, removed, or renamed require a new training run
- Accuracy degrades over time (model drift) — retraining restores performance
- You need to add a new language to an existing model
Retraining and new training are the same operation — each consumes credits according to the chosen mode and complexity.
Credit Guarantee
If a training run fails due to a system error on OTAI’s side, credits are refunded in full. Credits are only consumed for completed training runs.
Additionally, every subscription includes:
- 1 training run + 1 improvement iteration as part of the onboarding guarantee.
- If the agreed acceptance criteria are not met after the improvement iteration, the training fee is credited or refunded.