UplinkSwitch

    • Flagship AI models.
    Up to 70% cheaper than official list price.

    Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5, GPT-6.1 Sol and 77 more models through one key and one URL, every model between 40% and 70% under its developer's list price. Keep the SDK you already use. One prepaid balance, and volume pricing as your usage grows.

    See how it works Request a key

    https://api.uplinkswitch.com/v1 · 81 models and prices

    "Your"YourteamteamalreadyalreadyusesusesthetheOpenAI,OpenAI,AnthropicAnthropicororGeminiGeminiSDK.SDK.KeepKeepit.it.PointPointititatatoneoneURL,URL,handhandoutoutoneonekeykeyperperperson,person,paypayoneonebillbillatatwholesalewholesalerates,rates,andandseeseeeveryeverycallcallininoneoneplace."place."

    Works with the official OpenAI, Anthropic and Gemini SDKs, and with Claude Code. You change two settings: the URL and the API key.

    Claude
    OpenAI
    Gemini

    Everything your team needs around the model APIs. A key for each person, limits you control, a clear record of every request, and one wholesale rate card across all 81 models on the list.

    Low latency from anywhere. Your request enters UplinkSwitch at the Cloudflare location closest to you, gets checked against your key and limits, and goes straight down the managed route for that model.

    Every model in one list. Ask for the model list and you get everything your key can use: 81 models from Anthropic, OpenAI, Google, xAI, DeepSeek and 6 more labs. Each entry says who runs it, and the full price list is on the Models page.

    "Your key is fine, but the upstream isn't responding right now. Nothing was charged."

    What an error looks like

    Every error says what went wrong. You never pay for a request that didn't reach the model.

    See what every call cost. Each request is logged with the model, the route that ran it, the token counts the upstream reported, and the charge at the wholesale managed rate. Figures here are an example.

    Nothing hidden in the response. Four headers on every reply tell you the request id, which provider ran it, how it was billed, and where the token counts came from.

    Wholesale API pricing. Managed routes for 81 models from Anthropic, OpenAI, Google, xAI, DeepSeek and 6 more labs are priced against the developers' official list prices: never more than 60% of list, as low as 30% where our execution cost allows, and the same price whichever host answers. Every model and its rate per million tokens is on the Models page. Prefer your own provider account? Add your key and pay $0.001 per request on top of what the provider charges you. Open-weight models run on the same managed route at the same multiple, and teams with committed monthly volume get a per-account rate. Top-ups start at $10 and are not refundable. The chart is sample data.

    70%

    Up to, below official list price on output

    40%

    At least, below list on every metered model

    $0.001

    Per request, using your own provider key

    $10

    Minimum top-up

    Four routes, one rate card. Pick a model. UplinkSwitch sends the call down the managed route or to your own provider account, bills it at the published rate, and logs it the same way every time. The host is named on every model entry and response header; the whole list with prices is on the Models page.

    ANAnthropic models · managed route

    Claude

    Use the Anthropic SDK or Claude Code exactly as you do today. Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5. 13 models in all. Managed rate 40% to 70% under list, up to 70% on output. Or bring your own Anthropic key at $0.001 per request.

    OAOpenAI models · managed route

    GPT

    Chat, responses and embeddings through the OpenAI SDK. GPT-6.1 Sol, GPT-6 Sol, GPT-5.5. 15 models in all. Managed rate 40% to 70% under list, up to 70% on output. Or bring your own OpenAI key at $0.001 per request.

    GOGoogle models · managed route

    Gemini

    The Gemini SDK, pointed at one URL, streaming included. Gemini 3.8 Flash, Gemini 3.5 Flash, Gemini 3.1 Pro Preview. 14 models in all. Managed rate 40% to 70% under list, up to 40% on output. Or bring your own Google key at $0.001 per request.

    OWOpen-weight labs · managed route

    Open-weight models

    Grok 4.6, DeepSeek V4 Pro 0813, Kimi K3, Qwen3.7 Max and 35 more from xAI, DeepSeek, Moonshot AI, Qwen and others, on the same managed route and the same multiple. One wholesale rate per million tokens, volume pricing for committed teams.

    Common questions. Something else? Contact us.

    Get your key.

    Load $10, point your SDK at api.uplinkswitch.com, and your first request shows up in the log a second later. Volume pricing is arranged when you request a key.

    Request a key