EcoTrain API RaaS Pricing
Pay for measured compute savings
EcoTrain API uses a RaaS consumption model. Service fees are calculated from verified compute cost savings: fee deduction = compute cost saved x revenue-share ratio.
- Charged by actual savings
- Settled after training completion
- Consumption details available
- Monthly reconciliation supported
Billing formula
Prices are valid until September 30, 2026, 23:59.
Billing formula
Billing formula
EcoTrain API uses a RaaS consumption model. Service fees are calculated from verified compute cost savings: fee deduction = compute cost saved x revenue-share ratio.
Compute savings
Calculates the verified compute cost saved by a training task.
Service fee
The share ratio follows the task's measured saving tier.
Balance deduction
Final fees are settled after the training task completes.
Price categories
GPU-hour unit prices
Prices are in CNY per GPU-hour and are used to calculate saved compute cost.
| Category | Compute type | Unit price (CNY/GPU-hour) |
|---|---|---|
| Domestic accelerators | Ascend 910B2 | 2.08 |
| Domestic accelerators | Zhenwu 810E | 3.75 |
| Domestic accelerators | Moore Threads MTT S4000 | 1.40 |
| NVIDIA consumer | NVIDIA GTX 1080 Ti | 0.48 |
| NVIDIA consumer | NVIDIA RTX 2080Ti | 0.70 |
| NVIDIA consumer | NVIDIA RTX 3080 | 0.70 |
| NVIDIA consumer | NVIDIA RTX 3080 Ti | 0.81 |
| NVIDIA consumer | NVIDIA RTX 3090 | 0.85 |
| NVIDIA consumer | NVIDIA 4090D | 1.25 |
| NVIDIA consumer | NVIDIA 4090 | 1.25 |
| NVIDIA consumer | NVIDIA 5090D | 2.08 |
| NVIDIA consumer | NVIDIA 5090 | 2.08 |
| NVIDIA data center | NVIDIA V100 | 0.90 |
| NVIDIA data center | NVIDIA A800 | 3.80 |
| NVIDIA data center | NVIDIA A100-40G | 2.11 |
| NVIDIA data center | NVIDIA A100 | 4.80 |
| NVIDIA data center | NVIDIA H20 | 5.50 |
| NVIDIA data center | NVIDIA H800 | 8.12 |
| NVIDIA data center | NVIDIA H100 | 9.20 |
| NVIDIA data center | NVIDIA H200 | 10.60 |
| NVIDIA data center | NVIDIA B200 | 15.32 |
| NVIDIA data center | NVIDIA GB200 | 71.50 |
| NVIDIA data center | NVIDIA B300 | 17.84 |
| NVIDIA data center | NVIDIA A30 | 1.05 |
| NVIDIA data center | NVIDIA A40 | 2.05 |
| NVIDIA data center | NVIDIA L20 | 2.73 |
| NVIDIA data center | NVIDIA L40 | 2.73 |
| NVIDIA professional | NVIDIA RTX PRO 6000 | 5.27 |
| NVIDIA professional | NVIDIA RTX A4000 | 0.72 |
| NVIDIA professional | NVIDIA RTX A5000 | 1.53 |
| NVIDIA professional | NVIDIA RTX A6000 | 2.25 |
Settlement
Settlement and payment
EcoTrain uses prepaid balance. Recharge funds are used for acceleration service consumption.
Payment and settlement
Prepaid balance is used for acceleration services. Fee deduction = compute cost saved x revenue-share ratio, deducted from recharge balance or bonus balance.
Savings-based deduction
Fee deduction = compute cost saved x revenue-share ratio. Fees are deducted from recharge balance or bonus balance.
Bonuses and reminders
A single recharge above CNY 5,000 receives CNY 800 acceleration coupon credit. A single recharge from CNY 2,000 to CNY 5,000 receives CNY 200 acceleration coupon credit. A reminder is shown when the balance falls below CNY 500.
Additional notes
Rules for recharge, refunds, reconciliation, and invoices.
Ready to optimize training cost?
Connect EcoTrain API, pay for measured savings, and settle after each training task completes.