> For the complete documentation index, see [llms.txt](https://docs.stg.ai.neevcloud.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.stg.ai.neevcloud.com/gpu-ai-service/pricing-models.md).

# Pricing Models

NeevCloud offers flexible pricing models to match your usage patterns and budget requirements. Understanding these models helps you optimize costs while maintaining the compute power you need.

## On-Demand Pricing

### What is On-Demand?

On-demand instances are pay-as-you-go GPU resources. You pay only for the hours your instance runs, with no long-term commitment or upfront payment.

**How On-Demand Pricing Works:**

* Billing starts when your instance is running
* You're charged per hour (or per minute with hourly billing)
* You can start and delete instances anytime
* No minimum usage requirements
* No cancellation fees

### When to Use On-Demand

On-demand pricing is ideal when you:

* **Experiment with different configurations**: Testing various GPU types before committing
* **Run sporadic workloads**: Training jobs that happen irregularly
* **Need flexibility**: Projects with unpredictable duration
* **Develop and prototype**: Initial development before production deployment
* **Handle urgent tasks**: Sudden computational needs without waiting for reserved capacity

**On-Demand Pricing Example:** If you're training a model and you're not sure how long it will take:

* **Instance**: NVIDIA A100 (40GB VRAM)
* **Rate**: $2.50/hour
* **Training duration**: 6 hours
* **Total cost**: $2.50 × 6 = $15.00

You only pay for those 6 hours. If you stop the instance, billing stops immediately.

**Cost Management with On-Demand:**

* Set up monitoring alerts for long-running instances
* Stop instances when not actively computing
* Use automated shutdown scripts for completed jobs
* Monitor your usage dashboard regularly

## Reserved Instance Pricing

### What are Reserved Instances?

Reserved instances are GPU resources you commit to using for a specific period (typically 1 month, 3 months, or 12 months) in exchange for significant discounts.

**How Reserved Pricing Works:**

* You commit to a specific instance type for a set duration
* You pay upfront or monthly for your reservation
* Your hourly rate is significantly lower than on-demand
* The instance remains available to you during the reservation period
* You pay for the reservation whether you use it continuously or not

### Discount Levels

Typical savings compared to on-demand pricing:

* **1-month commitment**: \~10% discount
* **3-month commitment**: \~20% discount
* **12-month commitment**: \~35% discount

### When to Use Reserved Instances

Reserved instances make sense when you:

* **Run continuous workloads**: Training jobs that span days or weeks
* **Have predictable needs**: Regular batch processing or inference serving
* **Optimize costs for production**: Deployed models serving predictions 24/7
* **Conduct extended research**: Long-term research projects with steady GPU requirements
* **Require guaranteed availability**: Critical workloads that need assured capacity

**Reserved Pricing Example:** Compare costs for running an instance 24/7 for one month:

**On-Demand:**

* Instance: NVIDIA A100 (40GB VRAM)
* Rate: $2.50/hour
* Monthly hours: 730 hours
* **Total cost**: $2.50 × 730 = $1,825/month

**Reserved (1-month):**

* Discounted rate: $2.25/hour (10% savings)
* Monthly hours: 730 hours
* **Total cost**: $2.25 × 730 = $1,642.50/month

**Savings**: $182.50/month (10%)

For long-running workloads, reserved instances provide substantial savings.

**Important Considerations for Reserved Instances:**

* You commit to paying for the full reservation period
* If you stop using the instance early, you still pay the full commitment (we can refund the unused commitment when requested)
* The instance type and region are locked for the duration
* Plan your capacity needs carefully before committing

## Comparing Pricing Models

Use this decision framework:

**Choose On-Demand if:**

* Usage duration is uncertain
* You're experimenting or prototyping
* Workload is intermittent or unpredictable
* You need maximum flexibility
* Project may end suddenly

**Choose Reserved if:**

* You'll use the instance continuously for weeks/months
* Workload is predictable and steady
* Cost optimization is a priority
* You're deploying to production
* You have long-term compute requirements

## Billing and Cost Monitoring

### Understanding Your Bill

Your NeevCloud bill includes:

* Instance hours consumed (by instance type)
* Storage costs (for persistent volumes)

### Cost Tracking Tools

NeevCloud provides several tools to monitor spending:

* **Real-time dashboard**: See current running instances and hourly rates
* **Usage reports**: Historical usage broken down by instance and time period
* **Cost projections**: Estimated monthly costs based on current usage
* **Budget alerts**: Notifications (Email) when spending approaches your defined limits

### Best Practices for Cost Management

* **Tag your instances**: Use meaningful names and tags to track costs by project
* **Set up alerts**: Configure notifications for unusual spending patterns
* **Review regularly**: Check your dashboard weekly to identify idle resources
* **Automate shutdowns**: Use scripts to stop instances when jobs complete
* **Right-size instances**: Don't over-provision—choose appropriate VRAM and compute
* **Use reserved for base load**: Reserve capacity for steady workloads, use on-demand for peaks
* **Delete unused storage**: Clean up old snapshots and volumes you no longer need

## Pricing Transparency

NeevCloud displays pricing clearly during instance selection:

* Hourly on-demand rate shown for each instance
* Reserved pricing available in the reservation configuration panel
* Estimated monthly costs calculated based on 730 hours (24/7 usage)
* Regional pricing differences highlighted

You always know your costs before deployment, with no hidden fees or surprise charges.