Skip to content
runcrate
Console

Pricing

GPUs by the hour. Models by the token.

Named GPUs with a real shell, and 193 models behind one OpenAI-compatible API. One balance pays for both, at the live prices below.

Get an API keyTalk to an engineer

Billed by the second on most GPUs; a few bill by the minute.

Two real charges

October 19, 2026

  • A6000 instance

    Booted in 4 min 12 s, not billed. Then 7.9 s running, billed.

    $0.00143

    at $0.65/hr

  • DeepSeek-V3 call

    18 tokens in, 63 out.

    $0.0000649

    81 tokens

Total

$0.00149

Read from the billing ledger. A GPU's meter runs from running to deleted, by the second.

Plans

Pay as you go. Reserve when the job needs a cluster.

One prepaid balance pays for GPUs, models and storage. The prices on this page are what you are billed.

  • GPU instances

    For training, fine-tuning and serving

    from$0.36/ GPU-hour

    By the second on most GPUs, a few by the minute

    Booting is never billed

    Launch a GPU

    $20 buys one of

    • 5 hon one H100
    • 11 hon one A100 80GB
    • 22 hon one L40S

    Includes

    • SSH, Jupyter and VS Code in the browser
    • 22 GPU models on offer
    • Ready images for fine-tuning and serving

  • Inference

    For apps and agents calling models

    $0.273/ 1M input tokens

    $0.399 per 1M output, on DeepSeek V3.2

    Most images per image, video per second

    Get an API key

    $20 buys one of

    • 73Minput tokens on DeepSeek V3.2
    • 50Moutput tokens on DeepSeek V3.2
    • 761images from FLUX.1 Dev

    Includes

    • 193 models behind one API
    • OpenAI-compatible, so your client already works
    • No monthly fee

  • Storage volumes

    For datasets, weights and checkpoints

    $0.03/ GB a month

    Prepaid a week at a time while it exists

    Unused days refunded on delete

    Create a volume

    A month of

    • $3.00100 GB
    • $30.001 TB
    • $300.0010 TB

    Includes

    • Kept when you delete the instance
    • Mounted on the next instance you launch
    • 10 GB to 10,000 GB a volume

Reserved clusters. Prepay 1 to 12 months when you launch an instance, at any size. Want a custom quote? Talk to us.

Talk to an engineer

Inference

Open models, per 1M tokens. At the price you are billed.

One OpenAI-compatible API for every model. Input and output tokens are priced separately.

ModelTypePriceContextTry
Moonshot AI
Kimi K3 Fast
Moonshot AI$4.73in$23.63out$0.473cached input$4.73cache writeper 1M tokens
Text$4.73in$23.63out$0.473cached input$4.73cache writeper 1M tokens1MTry
DeepSeek
DeepSeek V4 Pro 0813
DeepSeek$1.37in$2.73out$0.0455cached inputper 1M tokens
Text$1.37in$2.73out$0.0455cached inputper 1M tokens1MTry
Meta
Llama 4 Scout 17B 16E
Meta$0.084in$0.252outno cache discountper 1M tokens
Text$0.084in$0.252outno cache discountper 1M tokens328KTry
Alibaba
Qwen3 Coder 480B A35B
Alibaba$0.315in$1.05out$0.105cached inputper 1M tokens
Text$0.315in$1.05out$0.105cached inputper 1M tokens262KTry
Alibaba
Qwen3 235B A22B 2507
Alibaba$0.0945in$0.578outno cache discountper 1M tokens
Text$0.0945in$0.578outno cache discountper 1M tokens262KTry
Z.ai
GLM 5.3 Prime
Z.ai$2.94in$9.24out$0.588cached input$2.94cache writeper 1M tokens
Text$2.94in$9.24out$0.588cached input$2.94cache writeper 1M tokens1MTry
OpenAI
gpt-oss Safeguard 20B
OpenAI$0.0788in$0.315out$0.0394cached input$0.0788cache writeper 1M tokens
Text$0.0788in$0.315out$0.0394cached input$0.0788cache writeper 1M tokens128KTry
MiniMax
MiniMax M3.1 Flash
MiniMax$0.158in$0.63out$0.0158cached input$0.158cache writeper 1M tokens
Text$0.158in$0.63out$0.0158cached input$0.158cache writeper 1M tokens1MTry
NVIDIA
Nemotron Nano 12B v2 VL BF16
NVIDIA$0.21in$0.63out$0.0315cached input$0.21cache writeper 1M tokens
Text$0.21in$0.63out$0.0315cached input$0.21cache writeper 1M tokens128KTry
Mistral AI
Mistral Large 3
Mistral AI$0.525in$1.58out$0.0525cached input$0.525cache writeper 1M tokens
Text$0.525in$1.58out$0.0525cached input$0.525cache writeper 1M tokens256KTry
See all 193 models

GPU compute

Every GPU we list, by the hour. The lowest live rate for each.

The same prices as JSON

GET /api/gpu-pricing

No key needed. Refreshed every 5 minutes.

Rate per

GPUMemoryFromAgainst the dearest GPUDeploy
B300BlackwellBlackwell · 288GB288GB$16.77/hrDeployDeploy
B200BlackwellBlackwell · 180GB180GB$7.66/hrDeployDeploy
H200Hopper · Billed by the minuteHopper · 141GB · Billed by the minute141GB$4.63/hrDeployDeploy
H100HopperHopper · 80GB80GB$3.55/hrDeployDeploy
GH200HopperHopper · 96GB96GB$2.98/hrDeployDeploy
A100 80GBAmpereAmpere · 80GB80GB$1.73/hrDeployDeploy
A100 40GBAmpereAmpere · 40GB40GB$1.99/hrDeployDeploy
L40SAda LovelaceAda Lovelace · 48GB48GB$0.89/hrDeployDeploy
L40Ada LovelaceAda Lovelace · 48GB48GB$1.30/hrDeployDeploy
RTX 5090BlackwellBlackwell · 32GB32GB$1.18/hrDeployDeploy
RTX 4090Ada LovelaceAda Lovelace · 24GB24GB$0.92/hrDeployDeploy
RTX 6000 AdaAda LovelaceAda Lovelace · 48GB48GB$1.06/hrDeployDeploy
RTX PRO 6000BlackwellBlackwell · 96GB96GB$2.45/hrDeployDeploy
A6000AmpereAmpere · 48GB48GB$0.57/hrDeployDeploy
A5000AmpereAmpere · 24GB24GB$1.85/hrDeployDeploy
A40AmpereAmpere · 48GB48GB$2.42/hrDeployDeploy
L4Ada LovelaceAda Lovelace · 24GB24GB$1.24/hrDeployDeploy
A4000AmpereAmpere · 16GB16GB$1.04/hrDeployDeploy
A10AmpereAmpere · 24GB24GB$1.68/hrDeployDeploy
V100 32GBVoltaVolta · 32GB32GB$0.36/hrDeployDeploy
V100 16GBVoltaVolta · 16GB16GB$0.36/hrDeployDeploy
GAUDI2architecture not listednot listed$1.56/hrDeployDeploy

22 GPU types, priced per GPU-hour. Billed by the second on most GPUs; a few bill by the minute. Prepay 1 to 12 months when you launch an instance, at any size. Want a custom quote? Talk to us.

What it costs to run

What real jobs cost. The four machines customers spent the most on.

From the billing ledger, September 19, 2026 to October 18, 2026: every launch by a paying customer that reached running and was settled, priced by what it was billed. Machines with at least 10 runs from 5 teams, each team weighing the same in the figures. Booting is never billed, so it is not in them.

Billing

How billing works. What you pay, and when.

How a dollar moves

01You add balanceBy card, before you use it
02Use draws it downGPUs by the second, models by the call
03Auto-recharge tops it upWhen it drops under your threshold

GPUs by the second

The meter starts when your instance is running and stops when you delete it. Booting is never billed. Most GPUs bill to the second; a few bill in whole minutes, rounded up, including about 30 seconds of teardown after you delete.

Models by the request

Each call is priced by its tokens, images or seconds of output at the prices on this page, and drawn from the same balance. No monthly fee and nothing to reserve.

Storage by the GB-month

A volume is $0.03 per GB a month, prepaid a week at a time while it exists. Delete it early and the unused days come back to your balance.

One prepaid balance

Add $20 to $10,000 at a time by card. It does not expire. A launch needs $10 or an hour of the machine, whichever is more, or just $10 with auto-recharge on.

Auto-recharge, with your numbers

Pick a threshold ($20 to $5,000) and a target ($50 to $10,000). Under the threshold, your card is charged back up to the target. A declined card turns it off, and we email you.

A warning before anything stops

Under $1.00 we try auto-recharge first, then email you and wait 15 minutes before running instances are deleted, or sooner once the balance reaches -$10.00. Each one is billed up to that moment.

Read the billing docs

Talk to an engineer

Tell us what you are running. We come back with a quote.

Pay as you go covers most work. For a reserved cluster or a large, steady inference volume, tell us the workload and we price it for the term.

  1. 01We reply within 24 hours.
  2. 02A call with an engineer to size the work.
  3. 03A quote for the GPUs, the volume and the term.
Prefer Slack?Create a Slack Connect channel
Which models do you need?

Every field is required unless marked optional.

FAQ

Questions about billing.