Batch AI compute from verified organizations
Submit inference jobs. We route them to underused GPUs from schools, labs, and companies. Transparent pricing. Carbon aware. Security first.
What is Coranor for customers?
Coranor is a compute marketplace that connects your batch AI workloads to underused GPUs. Instead of paying cloud premiums or managing your own hardware, you submit jobs and we handle the rest.
We start with verified organizations (schools, research labs, companies) because trust matters. Every provider is vetted. Every job runs in isolation.
This is not for realtime inference. It is for batch jobs: image generation, synthetic data, model evaluation, embedding pipelines. Jobs that can queue for a bit in exchange for lower costs and sustainable compute.
Best fit workloads
Coranor works best for batch, non-realtime inference.
Image Generation
Stable Diffusion, SDXL, and other diffusion models for batch image generation.
Synthetic Data
Generate training data at scale for ML models. Text, images, structured data.
Batch Inference
Process large datasets through LLMs or vision models. Not realtime, but reliable.
Model Evaluation
Run evaluation suites across model variants. Compare outputs systematically.
Reliability model
We are honest about where we are. This is early. We do not offer formal SLAs yet.
What we do offer: automatic retries, job queuing, progress tracking, and transparent status. If a node goes offline mid-job, we reschedule on another node.
As we scale and prove reliability, formal guarantees will follow. For now, we prioritize transparency. You will always know where your job is and what happened.
Current capabilities
- Job queuing with estimated wait times
- Automatic retries on failure
- Progress tracking and logs
- Upfront cost estimates
Coming later
- Formal SLAs with uptime guarantees
- Priority queues
- Dedicated capacity options
Simple, transparent pricing
No hidden fees. Know what you pay before you submit.
- Pay per GPU-hour. No minimums, no commitments.
- See pricing before you submit. No surprises.
- Queue priority options for time-sensitive work.
- Volume discounts for larger workloads (coming soon).
Exact rates depend on GPU type and availability. Current pricing starts around $0.30-0.50 per GPU-hour for consumer cards. Enterprise GPUs will be higher.
Frequently asked questions
What workloads are a good fit?
Batch inference jobs that can tolerate some queue time. Image generation, synthetic data creation, model evaluation, embedding generation. Not suitable for realtime inference or interactive applications.
How fast will my jobs run?
Depends on queue depth and job size. Typical batch jobs complete within minutes to hours. We show estimated completion time before you submit.
Is my data secure?
Yes. Jobs run in isolated containers. Data is encrypted in transit. We do not store your inputs or outputs longer than needed for job execution. See our Security page for details.
What GPUs are available?
Currently focusing on consumer and prosumer GPUs (RTX 3090, 4090, A6000). Enterprise GPUs coming as we onboard more verified providers.
How do I pay?
Credit card or invoicing for larger accounts. Billing is usage-based with transparent per-job costs shown upfront.
Can I get an SLA?
Not yet. We are in early access and prioritize transparency over promises we cannot guarantee. As we scale, formal SLAs will follow.
Join the waitlist
We are onboarding customers gradually. Sign up and we will reach out.
Note: This form will open your email client. If you prefer, you can email us directly at help@coranorlabs.com.