Visual-safety pricing
The open Hugging Face checkpoint has a $0 Apache-2.0 license fee. Falcons.ai separately sells shared-request API plans and a fixed-price self-hosted container.
Paid request pools are shared across Falcons.ai models. The effective figures below divide the monthly fee by the complete allowance; they are not separate per-image list prices.
| Plan | Price | Allowance | Cost basis |
|---|---|---|---|
| Free | $0 | 100 / model / month | Successful requests |
| Plus | $9.99 / month | 3,000 shared / month | ≈ $0.00333 at full use |
| Pro | $29.99 / month | 10,000 shared / month | ≈ $0.003 at full use |
| Ultra | $299.99 / month | 100,000 shared / month | ≈ $0.003 at full use |
| Containerized Self-Hosting | $5,000 / month | Unlimited API calls | Inside your deployment |
Source: Falcons.ai's live public billing configuration. Successful-request counting, video-frame use, retries, and plan overages should be confirmed before purchase.
Open checkpoint
Falconsai/nsfw_image_detection is public, ungated, and Apache-2.0 licensed. The roughly 343 MB ViT checkpoint has about 85.8 million parameters. Infrastructure, engineering, monitoring, moderation review, and security remain buyer costs.
Do not merge the routes
Falcons.ai exposes an authenticated NSFW endpoint, while Hugging Face maps the open checkpoint to HF Inference. Public documentation does not establish that Falcons.ai's paid API or enterprise container serves this exact repository revision. Benchmark each route separately.
Start with the open checkpoint on CPU. Measure latency and failure costs before renting a persistent GPU.
Compare the shared API pool against your successful-request volume and include review labor, retries, and data-handling requirements.
The $5,000 container only wins when utilization, privacy, or unlimited internal calls justify its fixed monthly cost.
Read the full download and pricing analysis and use our self-hosting cost guide before choosing infrastructure.
If testing proves that GPU inference is justified, check current RunPod GPU availability and pricing. AI Pricing Guru may earn a commission at no extra cost to you; it does not affect the pricing analysis.
Labs decision
The model is not placed in the text-only Cost-per-Task Leaderboard. A valid comparison needs a rights-cleared image suite, policy-specific labels, version-pinned preprocessing, false-positive and false-negative costs, latency, throughput, and infrastructure telemetry. See the documented image-harness blocker.