llmcloud.ai vs OpenRouter
OpenRouter proved the one-endpoint idea and built the market. We take a different position on three things: what the gateway costs, which providers get to serve you, and whether the product has an opinion.
| Dimension | llmcloud.ai | Typical marketplace gateway |
|---|---|---|
| Gateway fee | $0 on every tier. Tokens at provider cost; the platform is $10 per seat for teams that want org analytics and smart routing. | A percentage cut applied on top of inference spend. |
| Token pricing | Provider list price, passed through at cost, itemised on every receipt. | Provider price plus the platform cut, bundled into credits. |
| Unlimited subscriptions | Flat plans on our hosted fleet: $5 unlimited under 50B, $10 under 200B, and $20 that extends your Claude or OpenAI plan with free open-weight overflow when the lab caps you. | No equivalent — everything is metered from a prepaid credit balance. |
| Prepaid credits | No credit float. Pay for what you used, in arrears, in your own currency. | Prepaid credit balance held by the platform. |
| BYOK | Free on every tier. Your keys, zero markup, no per-request surcharge. | Available, typically with a fee on BYOK traffic. |
| Provider vetting | Published 5-criterion rubric, dated scorecards, monthly or quarterly re-evaluation. | Broad marketplace listing; vetting criteria not published. |
| Serving integrity | Weekly canary probes for quantization drift and silent model substitution. | Provider-reported; users detect drift from their own evals. |
| Model guidance | One named default, cheap alternate and escape hatch per workload. | Usage-ranked leaderboards; you decide. |
| Routing transparency | Every response returns the upstream, the policy that chose it and the cost delta. | Provider shown; policy rationale not exposed. |
| Compliance surface | Per-jurisdiction pages, residency pinning, attestation log with signed hashes. | Platform-level policies. |
| First-party hosting | We run open frontier and sovereign models in our own data centers, listed as an ordinary provider with no routing preference. | Pure marketplace — no owned inference capacity. |
| Self-host | Container image plus Terraform; control plane can run entirely in your VPC. | Hosted service only. |
Comparison reflects publicly documented behaviour of marketplace-style gateways at time of writing. Verify current terms with each vendor before switching.
The gateway fee is $0. Forever.
No routing cut, no credit float, no BYOK surcharge. You pay the provider's list price and see it itemised.
Every provider is graded.
A published rubric, our own probes, dated scorecards. Below the bar means out of default routing.
We name the model.
Per workload: one default, one cheap alternate, one escape hatch — and the directive that encodes it.
Frequently asked
If the gateway fee is $0, how do you make money?+
Seats and infrastructure, never a tax on routing: $10 per user per month for Team (advanced routing, org analytics, retained logs), Enterprise (SSO/SCIM, evidence bundles, dedicated capacity, SLA), self-host licences, and our own hosted fleet where we are the provider and earn the normal inference margin. The gateway fee on tokens stays $0 — that is the commitment.
Is the $0 gateway fee really forever?+
Yes. We never take a percentage of your token spend, and the Developer tier — full catalog, failover routing, BYOK, basic analytics — stays free. Paid tiers add platform capability, not a cut of your inference bill.
Are model IDs compatible with OpenRouter?+
Yes. provider/model IDs map 1:1, and :nitro / :floor variants translate to auto:speed and auto:cost. Migration is a base URL and a key.
Do you list every provider like OpenRouter does?+
No, and that is deliberate. A provider that will not publish retention terms, discloses no quantization, or fails the load harness does not serve default traffic. You can still pin it explicitly if you want it — you just have to opt in with the scorecard in front of you.