Pricing

Usage pricing for private AI compute.

Confidential VM

CVMs from

$0.06/hour

TDX-backed app servers for Docker workloads, APIs, agents, and proof-generating services.

Deploy CVM

GPU TEE

GPU TEE from

$4.40/hour

TEE-ready GPU capacity for private inference, training, and protected batch jobs.

Compare GPUs

Storage

CVM disk

$0.000139/GB/hour

Persistent storage is billed while the CVM exists, whether compute is running or stopped.

Ask about storage

Billing model

Compute stops. Storage stays until you delete it.

Pricing separates runtime from persistence. That makes short tests, demos, and one-off AI jobs cheaper without hiding disk costs.

Running

Compute$0.12/hr

Storage 100 GB$0.01/hr

Total~$0.13/hr

Stopped

Compute$0.00/hr

Storage 100 GB$0.01/hr

Total~$0.01/hr

Instance inventory

Choose the pricing lane.

tdx.small

Small TDX Instance

$0.06/hour

CPU1 vCPU

RAM2 GB

Disk20 GB

Deploy

tdx.medium

Medium TDX Instance

default

$0.12/hour

CPU2 vCPU

RAM4 GB

Disk20 GB

Deploy

tdx.large

Large TDX Instance

$0.23/hour

CPU4 vCPU

RAM8 GB

Disk20 GB

Deploy

View all

Enterprise clusters, reserved GPU slots, and custom network requirements are quoted through sales. All listed prices are usage estimates and can change with capacity availability.

Private AI model pricing

Pay per million tokens for private inference, or per image or audio minute for media models. Compare prices from the current model catalog.

Model	Context	Input / 1M	Output / 1M	Cache read / 1M	Image / audio
DeepSeek: DeepSeek V3.2	164K	$1.00	$1.00	$0.50	-
DeepSeek: DeepSeek V4 Flash 0731	1.0M	$0.44	$1.32	$0.028	-
DeepSeek: DeepSeek V4.1 Flash 1017	1.0M	$0.75	$1.60	$0.15	-
FLUX.2 klein 4B	-	-	-	-	$0.015 / image
Google: Gemma 4 31B	262K	$0.15	$0.46	$0.075	-
Jared Palmer: Kev 4B	8K	$0.042	Free	-	-
Meta: Llama 3.3 70B Instruct	131K	$2.00	$2.00	-	-
MoonshotAI: Kimi K2.6	262K	$1.09	$4.60	$0.37	-
MoonshotAI: Kimi K3	1.0M	$3.00	$15.00	$0.30	-
MoonshotAI: Kimi K3.1 Pro	1.0M	$2.00	$10.00	$0.20	-
OpenAI: GPT OSS 120B	131K	$0.15	$0.60	-	-
Phala: Gemma-4 26B-A4B Uncensored (Heretic)	262K	$0.15	$0.70	$0.05	-
Phala: Qwen3.8 27B Uncensored	1M	$0.30	$1.50	$0.05	-
Qwen: Qwen3 Embedding 8B	33K	$0.01	Free	-	-
Qwen: Qwen3 VL 30B A3B Instruct	128K	$0.20	$0.70	$0.03	-
Qwen: Qwen3.5 397B A17B	262K	$0.55	$3.50	$0.225	-
Qwen: Qwen3.5-27B	262K	$0.30	$2.40	$0.03	-
Qwen: Qwen3.6 27B	262K	$0.32	$3.25	$0.03	-
Qwen: Qwen3.6 35B A3B	262K	$0.20	$1.27	$0.056	-
Qwen: Qwen3.8 27B	1M	$0.20	$2.50	$0.05	-
Qwen2.5 7B Instruct	33K	$0.10	$0.20	-	-
Sentence Transformers: all-MiniLM-L6-v2	512	$0.005	Free	-	-
Voxtral Small 24B	-	-	-	-	$0.003 / min
Whisper Large v3	-	-	-	-	$0.006 / min
Z.ai: GLM 5.1	203K	$1.21	$4.20	$0.60	-
Z.ai: GLM 5.2	1.0M	$1.26	$3.00	$0.22	-
Z.ai: GLM 5.3	1.0M	$1.40	$4.40	$0.26	-
Z.ai: GLM 5.3 Flash	1.0M	$0.15	$0.50	$0.03	-

A dash means no cache-read rate is listed; cached input then uses the normal input rate. Caching depends on the provider and model. How prompt caching works

Billing questions

Common edge cases.

If I use a CVM for one week, do I pay the full month?

No. Compute is usage based. One week is billed for the active minutes used, and storage is prorated while the disk exists.

What happens when I stop a CVM?

Compute charges stop immediately. Storage charges continue so the disk can be preserved. Delete the CVM to stop both.

What welcome credit is available?

Some new workspaces are offered a welcome credit when they add a card, while the offer lasts. The amount and destination are set by the current Cloud offer. Balance pays for CVMs, GPU instances, and Private AI; gifted credits pay for CVMs and GPU instances.

What payment methods are supported?

Credit card through Stripe, crypto through Coinbase Commerce, and wire transfer for enterprise accounts.

Start with credits, then reserve what works.

Some new workspaces are offered a welcome credit when they add a card, while the offer lasts. Your Balance pays for CVMs, GPU instances, and Private AI. Check Cloud for the current offer details.

Start free trial Contact sales

Enterprise email: [email protected]