Infrastructure

Managed Models

The hosted BitRouter Cloud provider — one account, no upstream keys — the full model catalog with pricing, and automatic discounts on open models.

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

bitrouter auth login    # one-time device-flow sign-in
bitrouter start         # the `bitrouter` 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).

Supported models & pricing

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

ModelContextInput $/MOutput $/MProviders
anthropic/claude-fable-51M$10.00$50.001
anthropic/claude-haiku-4.5200k$1.00$5.001
anthropic/claude-opus-4.5200k$5.00$25.001
anthropic/claude-opus-4.6200k$5.00$25.002
anthropic/claude-opus-4.7200k$5.00$25.002
anthropic/claude-opus-4.81M$5.00$25.001
anthropic/claude-sonnet-4.5200k$3.00$15.001
anthropic/claude-sonnet-4.61M$3.00$15.003
deepseek/deepseek-v3.2131k$0.21$0.325
deepseek/deepseek-v4-flash262k$0.10$0.215
deepseek/deepseek-v4-pro256k$1.30$2.615
google/gemini-2.5-pro-preview1.0M$1.25$10.001
google/gemini-3-flash-preview1.0M$0.50$3.001
google/gemini-3-pro-preview1.0M$2.00$12.001
google/gemini-3.1-flash-lite-preview1M$0.25$1.501
google/gemini-3.1-pro-preview2M$2.00$12.002
google/gemini-3.5-flash1.0M$1.50$9.001
minimax/minimax-m2.5197k$0.23$0.906
minimax/minimax-m2.7197k$0.23$0.903
minimax/minimax-m31.0M$0.45$1.802
moonshotai/kimi-k2.5262k$0.44$2.005
moonshotai/kimi-k2.6256k$0.71$3.008
moonshotai/kimi-k2.7-code262k$0.71$3.001
openai/gpt-5.1400k$1.25$10.001
openai/gpt-5.1-codex400k$1.25$10.001
openai/gpt-5.1-codex-mini400k$0.25$2.001
openai/gpt-5.2400k$1.75$14.001
openai/gpt-5.2-codex400k$1.75$14.001
openai/gpt-5.3-codex400k$1.75$14.001
openai/gpt-5.4128k$2.50$15.003
openai/gpt-5.4-mini128k$0.75$4.503
openai/gpt-5.4-nano400k$0.20$1.251
openai/gpt-5.5128k$5.00$30.003
openai/gpt-oss-120b131k$0.25$0.691
qwen/qwen3.6-flash1M$0.19$1.131
qwen/qwen3.6-plus131k$0.38$2.254
qwen/qwen3.7-max1M$1.88$5.633
qwen/qwen3.7-plus1M$0.38$2.252
stepfun/step-3.5-flash262k$0.07$0.222
stepfun/step-3.7-flash256k$0.15$0.862
x-ai/grok-4.1-fast131k$0.20$0.501
x-ai/grok-4.20131k$1.25$2.502
x-ai/grok-4.31M$1.25$2.501
xiaomi/mimo-v2-flash66k$0.09$0.293
xiaomi/mimo-v2-omni262k$0.30$1.502
xiaomi/mimo-v2-pro1.0M$0.75$2.252
xiaomi/mimo-v2.5262k$0.30$1.503
xiaomi/mimo-v2.5-pro262k$0.75$2.253
z-ai/glm-5203k$0.75$2.406
z-ai/glm-5.1128k$1.05$3.307

Discounted open models

BitRouter 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.

25% off by default

Every model except the closed-source families — OpenAI (gpt-*), Anthropic (claude-*), Google (gemini-*), and xAI (grok-*) — is served by BitRouter'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

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

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 allowlists and 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

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

Reach out to set it up:

How is this guide?

On this page