$EX5 预售已开启 — 种子轮 · 500M 固定供应量 · 50% 销毁 · 通过 EX5 Exchange 完成 KYC参与 →
04 · EX5.AI · GPU MARKETPLACE + MODEL API

Rent GPUs. Run any model. Ship AI.

A GPU marketplace, a unified OpenAI-compatible model API, a fine-tuning pipeline and a chat workspace — one prepaid balance powers all of it. Email and card, no KYC.

EMAIL + CARDex5.aiCONSOLE · API · CLI
GPU MARKET · ON-DEMANDLIVE
H200 141GBus-ashburn-1 · 8 可用$1.37/hr租用 →
H100 SXM 80GBtr-istanbul-1 · 10 可用$1.08/hr租用 →
A100 SXM4 80GBtr-istanbul-1 · 2 可用$0.61/hr租用 →
L40S 48GBtr-ankara-1 · 5 可用$0.52/hr租用 →
RTX 5090 32GBpl-warsaw-1 · 14 可用$0.37/hr租用 →
Per-second billing · verified, benchmarked hosts打开 ex5.ai →
58
GPUs online right now
$0.33
per hour, starting price
50+
models behind one API
99.97%
API uptime, last 21 days

Four layers. One balance.

Mix raw compute, inference, fine-tuning and the chat workspace freely — everything is billed from the same prepaid credits.

01
GPU cloud

Bare-metal GPUs from verified hosts — H200 down to RTX 4090 — billed per second, cancel anytime.

02
Model API

One OpenAI-compatible endpoint for EX-1 and 50+ open models. Swap the base URL, keep your code.

03
Fine-tune & deploy

Upload a JSONL, train on spot GPUs, get a private autoscaling endpoint the moment training finishes.

04
chat.ex5.ai + CLI

A fast workspace for every model, and an agentic CLI for your terminal.

运作方式 · 所需条件

Load credits, pick a layer, ship.

No contracts, no cluster setup. Prepaid credits from $5, and you only pay for the seconds and tokens you use.

01
Load credits
Card or wire — prepaid credits in seconds, $5 minimum.
02
Pick your layer
Raw GPUs, the model API, a fine-tune job or the chat workspace.
03
Scale to zero
Per-second GPU billing, per-token inference, one invoice. Stop when you are done.
常见问题

常见问题

还不确定?客服支持全天候为您解答,请联系 support.ex5.com.

Yes. Point your existing client at ex5.ai/v1 with your EX5 key and keep your code — streaming and tool use included.
更多 EX5 产品 · 各有独立入口
立即开始
Start building on ex5.ai.
Create an account with an email, top up with a card, and rent your first GPU or call your first model in minutes.