Ghost Compute Console

// hardware mix

62 / 38

NVIDIA CC · AMD SEV-SNP

// avg utilization

74%

// median seal latency

182ms

p99 412ms

// fleet

Confidential GPU enclaves

Enclave	GPU	Region	Workload	Util	Last attestation	Status
gpu-h200-08af	H200 · CC	us-east	llama3.1-70b	91%	2s ago	sealed
gpu-h100-4421	H100 · CC	eu-west	mistral-large	68%	4s ago	sealed
gpu-mi300-9c1	MI300 · SEV	ap-south	FHE matcher	82%	1s ago	sealed
gpu-h200-1ee2	H200 · CC	us-west	vllm batch	54%	3s ago	sealed
gpu-h100-77b9	H100 · CC	eu-north	embeddings	33%	6s ago	re-attesting
gpu-mi300-22fd	MI300 · SEV	us-central	dark-pool	77%	2s ago	sealed

// live attestations

Streaming

[00:00:00] attest enclave gpu-h200-08af · sealed ok

[00:00:01] infer job 0x9b2…1c · llama3.1-70b

[00:00:01] fhe pool match · 14 orders · netted

[00:00:02] x402 402 → 200 · 0.0042 GHST

[00:00:02] solana attest tx 8K…Pq · slot 312884019

// quick actions

// hardware mix

62 / 38

NVIDIA CC · AMD SEV-SNP

// avg utilization

74%

// median seal latency

182ms

p99 412ms

// fleet

Confidential GPU enclaves

filter · h200all attested

Enclave	GPU	Region	Workload	Util	Last attestation	Status
gpu-h200-08af	H200 · CC	us-east	llama3.1-70b	91%	2s ago	sealed
gpu-h100-4421	H100 · CC	eu-west	mistral-large	68%	4s ago	sealed
gpu-mi300-9c1	MI300 · SEV	ap-south	FHE matcher	82%	1s ago	sealed
gpu-h200-1ee2	H200 · CC	us-west	vllm batch	54%	3s ago	sealed
gpu-h100-77b9	H100 · CC	eu-north	embeddings	33%	6s ago	re-attesting
gpu-mi300-22fd	MI300 · SEV	us-central	dark-pool	77%	2s ago	sealed

// vLLM · openai-compatible

x402 enabled

Private Inference endpoint

Prompts, system messages, KV caches and weights stay inside the enclave. The runtime returns only the response and a sealed attestation receipt.

# request curl https://api.ghostcompute.xyz/v1/chat/completions \ -H "x-payment: 402" \ -H "authorization: Bearer ghst_…" \ -d '{"model":"llama3.1-70b-cc","messages":[…]}' # response { "id":"chatcmpl_…","attestation":"0x9b2…1c","cost_ghst":0.0042 }

// served models

llama3.1-70b-cc8x H200
mistral-large-cc4x H100
qwen2.5-72b-cc8x H200
embed-ghst-v12x H100

// tps

14.8k

tokens / second · fleet

// p50 ttft

214ms

incl. enclave entry

// avg cost

0.0038

GHST · 1k tokens

// FHE matcher

encrypted

Confidential Dark Pool

Orders submitted as ciphertexts. The matcher runs entirely under FHE, settling net positions via Solana Confidential Balances. No mempool. No leakage.

// 24h volume

$48.2M

// orders matched

12,884

// match latency

1.8s

// slippage avg

3.1bp

// pool action

Submit sealed order

// pair

SOL / USDC

// side

// size · ciphertext

0xA9F…enc

// recent fills

Settlement log

Time	Pair	Notional (enc)	Fill	Attestation
14:08:21	SOL/USDC	0xA9F…enc	net	8K…Pq
14:08:14	JTO/USDC	0x71C…enc	net	4D…2w
14:07:58	BTC/USDC	0xF82…enc	net	9R…Hm
14:07:42	ETH/USDC	0x33B…enc	net	2Q…Lx

// attestation explorer

solana mainnet

Every job has a receipt.

Each enclave produces a hardware-signed quote. The Ghost runtime publishes a ZK-anchored summary on Solana. Public, verifiable, no account required.

// audits

OtterSecpassed
Trail of Bitspassed
Zellicpassed
Asymmetric Researchin flight

// latest attestations

Slot	Enclave	Job	Quote	Tx	Verified
312,884,019	gpu-h200-08af	infer · llama3.1-70b	0x9b2…1c	8K…Pq	ok
312,884,011	gpu-mi300-9c1	fhe · match	0x71C…aa	4D…2w	ok
312,884,004	gpu-h100-4421	infer · mistral-large	0xF82…30	9R…Hm	ok
312,883,996	gpu-h200-1ee2	vllm · batch	0x33B…7e	2Q…Lx	ok

// asset

GHST · Token-2022 on Solana

The work token of Ghost Compute. Operators stake GHST to earn job multipliers. Inference and dark-pool fees buy back and burn GHST. Confidential Balances keep treasury flows private.

// supply

1.00B

// circulating

218.6M

// staked

142.4M

// burned

3.82M

// distribution

Operators & staking40%
Ecosystem & grants22%
Treasury18%
Team (vested)12%
Liquidity8%

// utility

Stake to operate. Stake more for job priority and reward multipliers up to 2.4×.

// buyback & burn

40% of protocol revenue routes to weekly buyback, executed inside an enclave to prevent MEV.

// governance

Confidential voting via FHE quorum and outcome are public, ballots are not.

// 12-week rollout

From sealed enclave to public substrate.

// week 1 3 · enclave bring-up

NVIDIA CC H100/H200 and AMD SEV-SNP enclaves online. Attestation pipeline live. Ghost Worker alpha (Tauri).
// week 4 6 · private inference

vLLM packaged for CC. OpenAI-compatible API, x402 metering, GHST settlement. Internal audit by Trail of Bits begins.
// week 7 9 · dark pool

FHE matcher in production. Confidential Balances integration. OtterSec audit.
// week 10 12 · public substrate

Attestation Explorer public. GHST TGE. Zellic + Asymmetric Research final reports. Governance v1.

Confidential GPU enclaves

Confidential GPU enclaves

Private Inference endpoint

Confidential Dark Pool

Submit sealed order

Settlement log

Every job has a receipt.

GHST · Token-2022 on Solana

From sealed enclave to public substrate.

Connect a Solana wallet