Proof, not assertion
Live
What the InferAll gateway actually serves, right now. All counts are fetched server-side from /ai/v1/models and /ai/v1/health and refreshed at most once per hour — same numbers a developer can verify with curl.
Gateway status
Healthy
Models
122
Total
15
Free
107
Paid
“Free” means the model's pricing fields are explicitly $0 — for token models that's inputPerM and outputPerM both 0. Missing pricing metadata is not counted as free.
By vendor
| Vendor | Models | Free |
|---|---|---|
| openai | 27 | 1 |
| 26 | 1 | |
| other | 20 | — |
| anthropic | 14 | — |
| runway | 12 | — |
| nvidia | 8 | 8 |
| stability-ai | 2 | — |
| black-forest-labs | 2 | — |
| minimax | 2 | — |
| z-ai | 2 | 2 |
| meta | 2 | 2 |
| kwaivgi | 1 | — |
| sync | 1 | — |
| bytedance | 1 | — |
| devxpy | 1 | — |
| stepfun-ai | 1 | 1 |
Vendor is the namespace in the model id (e.g. meta/llama-… → meta). Many OSS models are routed through NVIDIA NIM regardless of vendor — see /docs for the routing details.
Beyond what's enumerated above: the gateway also passes any Replicate model id (owner/name) directly through to Replicate, and any NVIDIA NIM model id through to NIM — so the routable catalog is much larger than the number listed here. The counts on this page reflect what's explicitly enumerated in the registry today.
By source
59
Live (dynamic pricing pulled from upstream APIs)
63
Static (pricing pinned in the gateway registry)
Verify
Want to count for yourself? curl https://api.inferall.ai/ai/v1/models | jq 'keys | length' should equal 122 within the hour.