CommonCompute
Log inGet started
Log in
Get startedExplore workloads
Pricing

Quoted rate. Hard spend cap.

Every offered job returns a unit rate and estimated total before it runs. The unit rate is locked at submission; final cost uses measured usage and cannot exceed an explicit max_spend_usd cap. Failed jobs are never billed.

Featured workloads

Open-model workloads with a published comparison and a clear fit for the network. Reference prices are illustrative and may change; the unit rate in your Common Compute quote governs the run.

Workload
Unit
Batch
Priority
Compare against
Savings
Image upscale (4×)Preview
per image
$0.0033
$0.0066
ModelsLab Real-ESRGAN @ $0.0047
−30%
Transcription (Whisper)
per audio hour
$0.25
$0.50
OpenAI Whisper API @ $0.36
−30%
Subtitle generation
per video hour
$0.32
$0.63
AssemblyAI subtitles @ $0.45
−30%
Prices in USD. The API's realtime value selects priority dispatch at 2× batch; it does not include a latency or streaming SLO. Quoted unit rates are locked at submission; estimated totals can differ with measured usage and an explicit max_spend_usd is a hard cap. Failed tasks aren't billed.
Estimator

How much would your workload cost?

Workload
Priority
Priority uses the API's realtime value. It does not include a latency or streaming SLO.
Volume
10 1m tokens
Estimated cost
$0.63
Chat / LLM · Batch · 10 1m tokens
Unit price$0.063 / 1m tokens

Also available

Different hardware and hosted services may win on price or throughput for these workloads. Compare the current quote and measured fit for your own workflow before choosing a provider.

Workload
Unit
Batch
Priority
Compare against
Savings
Chat / LLM
per 1m tokens
$0.063
$0.13
Hosted LLM APIs (token-priced; models and input/output rates differ)
—
LoRA fine-tunePreview
per training minute
$0.055
$0.11
Cloud GPU rental (hardware-minute pricing; throughput and utilization differ)
—
Vision Q&A / doc understanding
per 1k images
$0.35
$0.70
Hosted multimodal APIs (token- or image-priced; resolution and quality differ)
—
Locked unit rate, optional max_spend cap, and no charges on failure. Reference comparisons are not guarantees of savings or performance.
FAQ

Pricing, explained.

How should I evaluate the price?
Pricing is workload-specific. Compare the published unit rate and your current quote with measured cost, quality, and throughput on your own workflow; reference comparisons are illustrative and do not guarantee savings.
Are quoted rates really locked?
Yes. The quoted unit rate is locked at submission. The displayed total is an estimate based on declared units; completion uses measured usage, and an explicit max_spend_usd remains a hard ceiling.
What counts as "batch" vs "priority"?
Batch and priority are queueing choices. Priority uses the API's realtime value and costs 2× the batch rate, but neither mode includes a dispatch-latency or streaming SLO.
Why can chat cost more than a hyperscale GPU service?
Dedicated GPU services can deliver higher transformer throughput at scale. Compare the current quote and availability for your own model and traffic pattern.
Do you charge for failed tasks?
No. Only successful, receipted results are billed. Retries are free.
Is there a free tier?
No free tier or trial. You add a card, buy credit ($10 minimum), and successful jobs draw against that balance based on measured usage at the quoted unit rate. Cost depends on the workload, model, priority, and usage; whatever you don't spend stays on your account.
Can I get volume pricing?
Your dashboard shows any rate available to your account. Contact sales for sustained-volume needs; reserved or dedicated capacity is not guaranteed.