Claude Haiku 4.5 Pricing on Fluxio: A First-Week Diary

§1 Monday, day one: opening a Fluxio diary with Claude Haiku 4.5

Here is a hypothetical first-week diary, written as an example rather than a customer story. You arrive at Fluxio with one goal in mind: run quick, low-cost streaming replies without guessing what a request will cost at the end. You pick Claude Haiku 4.5 as the model for the week, and you open the official Claude Code app, version 2.1.285 or newer, or the official Claude Desktop app at the same version or newer. Those are the only places Claude models run on Fluxio. When you connect an OpenAI-style client, it adds /v1 to the Fluxio API base, and that is the route the other model families use, but Claude Haiku 4.5 is not one of them.

Every new Fluxio account starts with $0.12 in credit, which is enough to send a few small requests and see the whole loop work: a prompt goes out, tokens come back one by one, and the dashboard shows what happened.

§2 Tuesday, day two: the pricing of Claude Haiku 4.5, line by line

Fluxio bills each request on the tokens the model reports. There is no flat per-request fee, and four token types carry separate prices. For Claude Haiku 4.5, per 1M tokens, input is $0.3458, output is $1.66, cache read is $0.0278, and cache write is $0.4153. The official list price for the same model is $1.00 input and $5.00 output per 1M tokens, so the Fluxio numbers sit well under that list price in both directions.

• – Input tokens at $0.3458 per 1M tokens cover the text you send.
• – Output tokens at $1.66 per 1M tokens cover the reply the model generates.
• – Cache reads at $0.0278 per 1M tokens cover cached context you reuse.
• – Cache writes at $0.4153 per 1M tokens cover context Fluxio stores for later reuse.

Because prices are per token rather than per seat, the daily number depends on your prompt lengths and on how long the answers run. That is why the usage log in Fluxio matters: it records each request, its tokens, and its cost, so you can read the pattern instead of estimating it.

§3 Wednesday, day three: what a dropped connection does to the bill

Mid-week, your laptop sleeps. The stream cuts off while the model is still writing. This is where the streaming-first design of Fluxio earns its place. If your client disconnects mid-stream, Fluxio keeps reading the reply to its end and bills only the tokens actually used, nothing more. There is no penalty added. Tokens that were already generated are not refunded, which is fair in both directions: you pay for what arrived, and you are not charged twice for work you never received.

The other side is worth planning around. A reply that was halfway finished still counts the tokens it burned, so an accidental retry is a second request on the meter. Fluxio reports each stream's real usage, and the dashboard keeps every request's token counts and cost, so a retry is visible rather than mysterious.

§4 Thursday, day four: the apps Claude Haiku 4.5 actually supports

This is the part people tend to mix up, so the diary keeps it plain. Claude Haiku 4.5 works only in the official Claude Code and Claude Desktop apps, version 2.1.285 or newer. It does not work in Cursor, Cherry Studio, Cline, OpenCode, any SDK, or cURL, and Fluxio does not pretend otherwise. Those OpenAI-format clients are where GPT, DeepSeek, Kimi, GLM, Grok, Qwen, Hunyuan, MiniMax, and MiMo models run. Gemini models run only through Gemini CLI on their native protocol. On Fluxio, replies arrive token by token over server-sent events on the OpenAI, Anthropic, and Gemini protocols, and every stream reports its real usage.

§5 Friday, day five: an example budget of 100 messages a day

For a concrete example, suppose you send 100 messages a day, each with 1,000 input tokens and 500 output tokens. At Claude Haiku 4.5 pricing, one such request costs $0.0003458 in input plus $0.00083 in output, or about $0.00118. Over seven days, 700 requests come to roughly $0.82. That is an example calculation, not a forecast, and your own prompt lengths will move it. Cache reads make repeated context cheaper at $0.0278 per 1M tokens, while a cache write of 10,000 tokens adds $0.004153.

Fluxio's pay-per-token balance carries no monthly fee. You can top up $1.39, $6.94, $13.89, $27.78, $69.44, or $138.89 through Stripe Checkout, which lets you size the experiment to the week rather than to a contract.

§6 Sunday, day seven: the balance page, the monthly plan, and where to ask

If you finish the week and want a larger runway, Fluxio also offers a Claude monthly plan at $62.36 for a $1100 allowance of official value over 30 days, capped at $350 a week, where each request draws its official-dollar value × 2.8. The plan covers Claude models only and uses its own API key. A larger version is $165.14 for a $2900 allowance over 30 days, at most $850 a week, with the same × 2.8 draw.

Top-ups stay flexible too. Unused balance from a top-up can be refunded within 4 days of that top-up, minus what was consumed and payment fees; you ask through the message page. Referrals pay 17% of what a directly invited friend pays and 5% for the next level, and rebates can move into the balance or be withdrawn from $12.50 by Alipay.

The account limits are modest and readable: 120 requests a minute per key and 8 streams open at once per account. The interface speaks six languages, English, Chinese, Spanish, Hindi, Portuguese, and French, and the usage log lists each request with its tokens and cost. If anything needs explaining, messages sent from the site contact page or the dashboard get replies in the dashboard; there is no support email to hunt for.

By Sunday, the diary reads like a ledger rather than a surprise. Claude Haiku 4.5 on Fluxio streams its replies token by token, prices every token type separately, and charges only for the work the model actually did, even when a connection drops halfway through.