Skip to content

Models & Providers

**Routeplane Cloud is on the Phase D roadmap and not yet shipping.** This page describes the future hosted provider and its planned pricing. Today, use [BYOK](/features/byok) with your own provider keys, or [local & private models](/integrations/models).

The Routeplane Cloud provider lets an agent call any model below with a single Routeplane account — no upstream provider keys, no per-provider signups. You pay Routeplane directly at the prices listed here, billed per request; failed requests aren’t billed.

Terminal window
routeplane cloud login # one-time sign-in (available once Cloud ships)
routeplane start # the `routeplane` provider auto-enables once signed in

Prefer your own provider accounts? Use BYOK instead — you pay providers directly at their list price. Running your own model? See local & private models (free).

Prices are USD per million tokens, refreshed continuously from the live catalog. Open models are served 25% below official by default — see Discounted open models below.

Every model above is served by one or more registered providers. Membership lives in the public, open-source provider-registry — anyone can register a provider. The list refreshes from the registry continuously, so a newly-merged provider shows up within minutes.

Routeplane runs its own self-hosted provider for open models, priced 25% below official rates. You get that price automatically — and open-source builders can apply for a deeper custom discount.

Every model except the closed-source families — OpenAI (gpt-*), Anthropic (claude-*), Google (gemini-*), and xAI (grok-*) — is served by Routeplane’s self-hosted provider at 25% below the model’s official price.

This takes no suffix and no configuration. Because the self-hosted provider is the cheapest source for these models, normal routing already sends your requests there and bills the discounted rate. (The four closed-source families above aren’t on the self-hosted provider, so they route to their usual upstreams at standard pricing.)

Pin to the self-hosted provider with :discount

Section titled “Pin to the self-hosted provider with :discount”

Append :discount to a model id to route the request specifically to Routeplane’s self-hosted provider:

Terminal window
curl http://127.0.0.1:4356/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "moonshotai/kimi-k2.6:discount",
"messages": [{"role": "user", "content": "Translate to French: Hello."}]
}'

The suffix rides on the model string — no body fields, no SDK — and works the same on the OpenAI, Anthropic, and Google surfaces (/v1/messages, /v1beta/models/{model}:generateContent). Use it to guarantee your traffic lands on the discounted self-hosted supply; it’s also where any custom discount on your account applies.

`:discount` never changes authorization. [Guardrail](/features/guardrails) allowlists and [BYOK](/features/byok) rules judge `moonshotai/kimi-k2.6:discount` exactly as `moonshotai/kimi-k2.6` — the suffix can't widen or bypass a policy.

Custom discounts up to 50% for open-source projects

Section titled “Custom discounts up to 50% for open-source projects”

Building an open-source agent harness or another open-source project on Routeplane? We offer customized discounts — up to 50% off — for you and your community.

Reach out by opening a GitHub issue on the routeplane repository once Cloud ships.