Pricing

Pay how you want.
Build without limits.

Stay free, or top up credits as you build.

Every plan includes the every major AI provider, same low-latency edge, same unified API key. For one-time top-up credits, see consumer pricing.

API calls: upstream cost × 1.2, charged in tokens (1 token = $0.0001 upstream cost). One balance across Chat, Image, Video, Mentor, and the API.

Accepted:
VISA
Mastercard
Alipay
Free
No card
$0
try before you commit
  • Generous daily request limit · resets midnight UTC
  • GPT-5 mini, Claude Haiku, Gemini Flash, DeepSeek
  • low-latency edge routing
  • 60-second response cache
  • Auto-fallback to free models when upstream errors
  • Frontier models (GPT-5.2, Opus, Gemini Pro)
  • SLA & priority support
Start free →
Pay-as-you-go

Top up credits. Use them when you build.

No subscription on the developer side. Buy credits in any amount ($5 minimum), drawn from your balance as you call the API. One-time Stripe charge — cancel-anytime by simply not topping up again.

Failover, not failure.

When the primary provider errors or rate-limits, Aperture automatically falls back to a high-quality free model (GPT-OSS 120B, Nemotron 120B) so your workflow keeps moving. When your credits run out you'll see a clear 402 — no surprise charges, no silent downgrade.

Questions worth asking

Do I bring my own OpenAI / Anthropic API key?

No. Aperture maintains its own upstream relationships with each provider. You use one Aperture key (sg-...) and we route to whichever model your code requests.

Subscription or credits — when do I pick which?

Pro subscription is best when you ship something that calls models every day. Credits suit weekend projects, prototyping, or unpredictable spikes. You can run both at once: Pro covers your daily 10K, credits soak up the overflow.

Do credits expire?

No. Once you top up, credits stay in your account until you burn them down. No 6-month or 12-month clocks.

What payment methods work for Asian users?

Visa, Mastercard, and Amex via Stripe. Most cards issued in Asia work. Local payment rails (e-wallets) are not currently supported — Stripe card is the only payment path.

Where does my data go?

Requests pass through our low-latency edge (DigitalOcean SGP1) on the way to the upstream provider. We do not log prompt or response bodies — only metadata (model, token count, status, latency) for billing and analytics. Cache entries are keyed on input hash and live in Redis with the TTL of your plan.

What's the rate limit and concurrency?

Each plan ships with a generous daily request quota that resets at midnight UTC. Over-quota requests return HTTP 429 daily_quota_exceeded with the exact reset time. Upstream provider rate limits also apply (you'll see those as the original provider 429s, pass-through).

Can I cancel or refund?

Subscriptions cancel anytime via Stripe portal — your plan stays active through the end of the billing month, no charge after. Credits are non-refundable once spent, but unused credits refund on request within 14 days of purchase.

What if a provider goes down?

Aperture exposes a single /api/v1 endpoint. Switch models by changing the model name in your request — no path changes, no new keys.

Why Singapore and not Tokyo / Mumbai?

Singapore peers excellently with Indonesia, Vietnam, the Philippines, Malaysia, Thailand — and still keeps sub-100ms hops to Tokyo, Sydney, and the US west coast. For an SEA-first audience it gives the best blended latency without splitting infrastructure across regions.

Can I see usage in real time?

Yes. You get a dashboard at /developers/dashboard.html with your API key, current plan, and code snippets. Detailed usage stats coming next.

Start free, top up when you need more.

Either way, your API key is live in under 30 seconds.