One balance, everything
Agents, models, GPU hosting, compute, and search all draw from the same prepaid credits. 1 credit = $0.001.
Back to overview
usage-based credits
Usage-based pricing. Prepay credits, pay for what you use.
One prepaid balance covers coding agents, model routing, GPU hosting, compute, and search. Pay by token or second, with optional plans for bundled credits and guaranteed machines.
app credits buy --packs 25What you get
Simple credits
Pay as you go
Search pricing
Model hosting
Spend caps
Metered, with caps
Pay per token and per second. Set spend caps per project, key, app, or agent run so there are no surprises.
Plans are optional
Pro, Ultra & Max bundle monthly credits at a better rate plus guaranteed machines — usage beyond them just continues at list price.
Start free, or save with a plan
Plans add monthly credits at a better rate, guaranteed machines, and priority in the queue. Same balance — go over your plan and you just keep paying the usage rates below. No subscription required. Cancel anytime.
Free
$0/mo
Start free. Pay only for what you use.
- Prepaid credits, no minimum
- Shared worker pool
- Every API + the CLI
- Web, paper & agentic search
Pro
$20/mo
For solo builders shipping every day.
25,000 credits / month
Save 20% vs pay-as-you-go
- 25,000 credits / month
- 1 guaranteed agent machine
- RA1 + papers + agentic search quota
- Priority over free tier
Ultra
Popular$60/mo
For teams running agents in parallel.
80,000 credits / month
Save 25% vs pay-as-you-go
- 80,000 credits / month
- 3 guaranteed machines + priority queue
- Higher first-party product quotas
- Best value per credit
Max
$200/mo
For heavy, always-on workloads.
280,000 credits / month
Save 29% vs pay-as-you-go
- 280,000 credits / month
- 6 guaranteed machines + top priority
- Largest RA1 / papers / search quotas
- First-party credits across products
Or pay as you go — one prepaid balance
Everything on app.nz — coding agents, the model gateway, GPU model hosting, rented CPU/GPU compute, and web & paper search — is usage-based and billed from one prepaid credit balance. Top up once and use anything. No subscription required.
Prepay credits
Top up once. 1 credit = $0.001. No subscription required, no minimums, no monthly commitment.
Use anything
Agents, the model gateway, GPU model hosting, rented compute, and search all draw from the same balance.
Pay for what you use
Metered to the token and to the second. Set spend caps per project, key, app, or agent run.
Models
Per 1M tokens, through one OpenAI-compatible gateway. Auto routes pick the right model for the job.
| Model | Provider | Input / 1M | Output / 1M |
|---|---|---|---|
| openpaths/auto-fast | DeepSeek | $0.168 | $0.336 |
| openpaths/auto-cheap | OpenAI | $0.240 | $1.50 |
| openpaths/auto | Google Gemini | $1.80 | $10.80 |
| openpaths/auto-code | OpenAI | $6.00 | $36.00 |
| openpaths/auto-reasoning | OpenAI | $0.900 | $5.40 |
| claude-opus-4-8 | Anthropic | $6.00 | $30.00 |
| claude-sonnet-5 | Anthropic | $3.60 | $18.00 |
| claude-haiku-4-5-20251001 | Anthropic | $1.20 | $6.00 |
| gpt-5.4 | OpenAI | $3.00 | $18.00 |
| gpt-5-mini | OpenAI | $0.300 | $2.40 |
| gemini-2.5-pro | Google Gemini | $1.50 | $12.00 |
| gemini-3.5-flash | Google Gemini | $1.80 | $10.80 |
| deepseek-chat | DeepSeek | $0.336 | $0.504 |
| grok-4.5 | xAI | $2.40 | $7.20 |
| grok-4.3 | xAI | $1.50 | $3.00 |
| llama-3.3-70b-versatile | Groq | $0.708 | $0.948 |
A representative slice — see every model and provider.
Hardware pricing
Billed per second of wall-clock, from the same balance. GPU model endpoints scale to zero, so idle hardware costs nothing — and every machine is priced below Replicate's rate for the same GPU.
| Hardware | Price | GPU | CPU | GPU RAM | RAM |
|---|---|---|---|---|---|
| CPU Smallcpx21 | $0.000003/sec$0.010/hr | — | 3x | — | 4GB |
| CPU Mediumcpx32 | $0.000005/sec$0.018/hr | — | 4x | — | 8GB |
| CPU Largecpx41 | $0.000010/sec$0.038/hr | — | 8x | — | 16GB |
| Windows CPU Mediumwin-cpx32 | $0.000005/sec$0.018/hr | — | 4x | — | 8GB |
| Windows CPU Largewin-cpx41 | $0.000010/sec$0.038/hr | — | 8x | — | 16GB |
| A100 80GBgpu-a100 | $0.000729/sec$2.624/hr | 1x GPU | 12x | 80GB | 80GB |
| H100 80GBgpu-h100 | $0.001329/sec$4.784/hr | 1x GPU | 16x | 80GB | 80GB |
| H200 141GBgpu-h200 | $0.001773/sec$6.384/hr | 1x GPU | 16x | 141GB | 141GB |
| A40 48GBgpu-a40 | $0.000178/sec$0.640/hr | 1x GPU | 9x | 48GB | 48GB |
| L40Sgpu-l40s | $0.000382/sec$1.376/hr | 1x GPU | 8x | 48GB | 48GB |
| RTX 3090 24GBgpu-rtx3090 | $0.000098/sec$0.352/hr | 1x GPU | 8x | 24GB | 24GB |
| RTX 4090 24GBgpu-rtx4090 | $0.000151/sec$0.544/hr | 1x GPU | 8x | 24GB | 24GB |
| T4 16GBgpu-t4 | $0.000084/sec$0.304/hr | 1x GPU | 4x | 16GB | 16GB |
| TPU v5ecoming soontpu-v5e | $0.000640/sec$2.304/hr | 1x TPU | 24x | 16GB | 48GB |
| TPU v6e (Trillium)coming soontpu-v6e | $0.001440/sec$5.184/hr | 1x TPU | 44x | 32GB | 176GB |
You only pay while a machine is running your work, metered per second. TPUs land with the Google Cloud integration — pricing shown so you can plan; Google-Cloud-only hardware carries GCP's higher list prices.
Search
Flat usage pricing for live web and academic search, drawn from the same balance.
Web search
$7.70 / 1,000 requestsLive web results with citations
Paper search
$1.00 / 1,000 requestsAcademic + preprint search
Media optimization
Cloudflare-style responsive images and AV1 video ladders, saved into your artifact filesystem.
Responsive image set
set up to 8 generated variants
Dynamic reads are cacheable; generated artifact variants are billed once.
AV1 video ladder
one minute of encoder time at the default GPU rate
Uses AV1 NVENC when available, then CPU AV1/VP9/H.264 fallback.