Claude relay service tips: 5 ways to get more from each Fluxio top-up

§1 Point Claude Desktop 2.1.285 or newer at the Fluxio API base

Claude models on Fluxio run inside the official Claude Desktop app and the official Claude Code app, version 2.1.285 or newer. That version floor is the whole rule: on Fluxio, Claude models do not run in Cursor, Cherry Studio, Cline, OpenCode, any SDK or cURL, so there is little sense trying to force them into a client that was never supported for them.

Fluxio's API base is api.fluxioapi.com. OpenAI-style clients add /v1 to that base, which matters when you pair Fluxio with GPT, DeepSeek, Kimi, GLM, Grok, Qwen, Hunyuan, MiniMax and MiMo models, or when Codex runs GPT models. Claude sits in a different lane entirely, and the official apps are where it belongs.

Once the app is pointed at Fluxio, each request is billed on the tokens the model reports. Fluxio prices four token types separately — input, output, cache write and cache read — and adds no flat per-request fee. Short, focused exchanges therefore cost less than sprawling ones, and the usage log in the Fluxio dashboard shows every request with its tokens and cost, so you can see exactly where a top-up went.

§2 Match the Claude model to the shape of your prompts

Fluxio lists Claude Fable 5, Claude Fable 5.1, Claude Haiku 4.5, Claude Opus 5.5, Claude Sonnet 5 and Claude Sonnet 5.5, each with its own per-1M-token rates. Here is the Fluxio side of the ledger:

• – Claude Fable 5: input $3.50, output $17.50, cache read $0.2694 and cache write $4.31 per 1M tokens, against an official list price of $10.00 input, $50.00 output and $1.00 cache read.
• – Claude Fable 5.1: input $3.88, output $19.40, cache read $0.2319 and cache write $4.78 per 1M tokens, against an official list price of $10.00 input, $50.00 output and $0.25 cache read.
• – Claude Haiku 4.5: input $0.3458, output $1.66, cache read $0.0278 and cache write $0.4153 per 1M tokens, against an official list price of $1.00 input, $5.00 output and $0.10 cache read.
• – Claude Opus 5.5: input $1.51, output $7.13, cache read $0.0681 and cache write $1.78 per 1M tokens, against an official list price of $4.00 input, $20.00 output and $0.20 cache read.
• – Claude Sonnet 5: input $1.21, output $6.04, cache read $0.1056 and cache write $1.51 per 1M tokens, against an official list price of $2.00 input, $10.00 output and $0.20 cache read.
• – Claude Sonnet 5.5: input $1.19, output $5.97, cache read $0.0917 and cache write $1.49 per 1M tokens, against an official list price of $2.00 input, $10.00 output and $0.10 cache read.

A worked example, to show how the figures behave: if you send 100 messages a day, each carrying 1,000 input tokens and 500 output tokens, Claude Sonnet 5.5 on Fluxio costs about $0.4175 a day, or roughly $12.53 across 30 days. The same 100-message pattern on Claude Haiku 4.5 runs at about $0.1176 a day, or roughly $3.53 across 30 days. That is an example based on those token counts, not a forecast of your own traffic, but it shows why the model you choose often matters more than the size of the top-up behind it.

§3 Let Fluxio stream token by token and survive a dropped connection

Fluxio is streaming-first: replies arrive token by token over server-sent events on the OpenAI, Anthropic and Gemini protocols, and every stream reports its real usage as it runs. Inside Claude Desktop that means you watch the answer form instead of waiting for a sealed response.

The dropped-connection policy is worth reading twice. If your client disconnects mid-stream, Fluxio still reads the reply to its end and bills only the tokens actually used. No penalty is added. Tokens that were already generated are not refunded, which is a good reason to let a stream finish rather than cancelling mid-answer.

Two limits shape how much you can run in parallel on Fluxio: 120 requests a minute per key, and 8 streams open at once per account. When a heavy session approaches the stream ceiling, closing finished streams is usually enough.

§4 Use cache reads to make repeated context cost less

The fourth token type is where careful habits pay off. On Fluxio, a cache write is billed separately from a cache read, and for Claude Fable 5 the two sit at $4.31 and $0.2694 per 1M tokens. For Claude Fable 5.1 they sit at $4.78 and $0.2319. Claude Haiku 4.5 charges $0.4153 for a cache write and $0.0278 for a cache read; Claude Opus 5.5 charges $1.78 and $0.0681; Claude Sonnet 5 charges $1.51 and $0.1056; Claude Sonnet 5.5 charges $1.49 and $0.0917.

The pattern holds across the Claude line: writing context once and reading it back many times costs far less than re-sending that context as fresh input every turn. If your Claude Desktop sessions reuse the same documents, project notes or system instructions, the cache entries in the Fluxio usage log show whether that reuse is being billed at the cache-read rate.

Compare the two Fable entries carefully, since they are easy to mix up: Fable 5 reads cache at $0.2694 while Fable 5.1 reads it at $0.2319, even though Fable 5.1's input and output rates are higher. Whether that trade works for you depends on how much of your traffic is cache reads rather than fresh input.

§5 Top up in measured steps and watch the 4-day refund window

Fluxio pay-per-token has no monthly fee. Top-ups run through Stripe Checkout at $1.39, $6.94, $13.89, $27.78, $69.44 and $138.89, so you can size a purchase to the pace of your Claude Desktop usage instead of committing to one large block. Every new account also receives $0.12 in sign-up credit, enough to cover a little of the first session.

The refund rule is straightforward: unused balance from a top-up can be refunded within 4 days of that top-up, minus what was consumed and payment fees, and you ask through the message page. That makes a modest first top-up a sensible way to learn the rhythm of your own usage before scaling up.

Fluxio also runs a referral scheme: 17% of what a directly invited friend pays, then 5% at the next level. Rebates can move into your balance, or be withdrawn from $12.50 by Alipay. Support requests go through the site contact page or the dashboard, and replies appear in the dashboard, with no support email to hunt for. The interface speaks six languages: English, Chinese, Spanish, Hindi, Portuguese and French.

§6 Compare the Claude monthly plans with pay-per-token

For steady Claude traffic, the monthly plans on Fluxio work with your own API key and cover Claude models only. The first is $62.36 for an allowance of up to $1100 over 30 days, at most $350 a week. The second is $165.14 for an allowance of up to $2900 over 30 days, at most $850 a week. Allowances on both are counted at the plan rate, not at list prices, with multiplier × 2.8.

That distinction matters when you read the price lists above. The official list prices are reference points for what the same tokens would cost elsewhere; what actually draws down a Fluxio plan allowance is the plan rate. If your Claude Desktop habit is heavy enough that a weekly cap of $350 or $850 feels comfortable, the plan route gives you a predictable ceiling. If your usage is spiky, or you are still calibrating, the pay-per-token balance with no monthly fee keeps spending proportional to what you actually send.

Whichever route you take, Fluxio's streaming-first delivery, per-token billing and per-request usage log make the cost of a Claude session something you can watch rather than guess.