Plans for your Supercomputer
One subscription runs your personal agent, its machine, and a monthly credit allowance for renders. Credits are one pool — chat and API spend the same balance.
Cancel anytime · billed via Paddle
Starter
For first-time AI content creators
$19
/mo
billed annually
270 credits / month
- ✓ 2 parallel videos · 4 parallel images
- — Selected models only
- — Early access to advanced features
- — Lowest cost per credit
- ✓ Supercomputer storage: 100 MB
- ✓ Efficient supercomputer
- — Smart supercomputer — no access
- — Scheduled jobs — no access
Checkout for this plan isn't wired up yet
Plus
For consistent and easy AI content creation
$31
$59
/mo
billed annually
1,200 credits / month
- ✓ 6 parallel videos · 8 parallel images
- ✓ All models & features
- ✓ Early access to advanced features
- — Lowest cost per credit
- ✓ Supercomputer storage: 2 GB
- ✓ Efficient supercomputer
- ✓ Smart supercomputer
- ✓ Scheduled jobs: up to 2
Checkout for this plan isn't wired up yet
Best value
Ultra
For creators building AI projects
$63
$129
/mo
billed annually
3,000 credits / month
price scales pro-rata with credits
- ✓ 8 parallel videos · 8 parallel images
- ✓ All models & features
- ✓ Early access to advanced features
- ✓ Lowest cost per credit
- ✓ Supercomputer storage: 5 GB
- ✓ Efficient supercomputer
- ✓ Smart supercomputer
- ✓ Scheduled jobs: up to 10
Checkout for this plan isn't wired up yet
Model access per plan
Rendered from the live catalog — the same table your agent sees via
estimate_cost. Starter runs the selected set; Plus and Ultra run
everything.
| Model | Type | Credits | Starter | Plus / Ultra |
|---|---|---|---|---|
| Nano Banana | image | 1 | ✓ | ✓ |
| Seedream | image | 1 | ✓ | ✓ |
| Flux Pro | image | 2 | ✓ | ✓ |
| GPT Image 2 | image | 3 | — | ✓ |
| Nano Banana Pro | image | 2 | ✓ | ✓ |
| Flux Max | image | 3 | — | ✓ |
| Quick Video | video | 1/s | ✓ | ✓ |
| Cinematic Video | video | 4/s | — | ✓ |
| Veo 3.1 Fast (audio) | video | 2/s | ✓ | ✓ |
| Veo 3.1 (audio) | video | 6/s | — | ✓ |
Pass model explicitly or let the router pick. estimate_cost
gives the exact number before a real request.
Building with the API / MCP instead?
Keys, 25 free credits and pay-as-you-go top-ups don't need a subscription — the plans above gate the web chat product, not the MCP surface. One account, two doors; credits are one pool.
Get an API key