Cherry Studio API Relay Pricing: Quanta's $1.39 Top-Up

01 · How Quanta bills your first API call

Quanta is built for teams running heavy, sustained workloads, and its billing model starts where most relays leave you guessing. You open an account, receive $0.50 in trial credit, add funds, and make calls. There is no monthly fee and no minimum spend. Every request is metered on the real tokens it consumes, and the dashboard lists the tokens and the charge for each individual call.

That is the whole structure: a prepaid balance, published prices, and a record of what each call actually cost. Quanta does not hide the mechanics behind subscription tiers or bundled allowances. If your team sends a large prompt, receives a long answer, or writes to cache, each of those actions is metered and shown separately.

For teams, the important point is that the numbers are visible before and after the work runs. You can see how a balance is being spent at the call level, which makes it practical to connect usage to a specific application, key, or workload without relying on estimates.

02 · What the four token prices actually mean

Quanta prices per 1M tokens and separates four categories: input, output, cache write, and cache read. This matters because a workload that looks identical on the surface can cost very different amounts depending on how the prompt is structured and whether cached context is being reused.

• › Input is the text your team sends to the model.
• › Output is the text the model returns.
• › Cache writes are charged when context is written to cache.
• › Cache reads are charged when cached context is reused.

Published prices include, for example, Claude Opus 5.5 (claude-opus-5-5) at $1.79 / $8.46 / $2.11 / $0.0819, Claude Sonnet 5.5 (claude-sonnet-5-5) at $1.25 / $6.28 / $1.57 / $0.1097, GPT-6.1 Sol (gpt-6.1-sol) at $0.2083 / $1.01 / - / $0.0403, and Gemini 2.5 Flash (gemini-2.5-flash) at $0.2361 / $1.89 / $0.2361 / $0.0236. These are approximate USD figures; charges are made in CNY. Where cache writes are not offered for a model, the price is shown as -.

Here is a simple example, not a forecast: at a 3:1 input-to-output ratio with no cache hits, a team running 10 million tokens a month on Claude Opus 5.5 would spend about $34.58, and 100 million tokens about $345.83. The same example on Claude Sonnet 5.5 comes to roughly $25.07 and $250.69, while GPT-6.1 Sol comes to about $4.10 and $40.97, and Gemini 2.5 Flash to about $6.49 and $64.93. Model choice, prompt length, and caching behaviour all move the final figure, which is exactly why Quanta itemises each call.

03 · Prepaid balances, top-ups, referrals, and refunds

Because Quanta runs on a prepaid balance, funding is straightforward. The available top-up amounts are $1.39, $6.94, $13.89, $27.78, $69.44, and $138.89. A smaller top-up suits a team testing a new integration; larger amounts suit production traffic that runs continuously. There is no minimum spend, so you are not committed to a volume you have not used yet.

Refunds are also defined rather than discretionary. Unused balance can be refunded within 15 days of the top-up, which gives a team a clear window to evaluate a workload before the balance is considered spent.

Quanta also publishes its referral terms: referred accounts generate a reward of 19% of top-ups, and 5% one level further. Rewards can be moved into the balance, or withdrawn from $27.78. For teams already paying for API capacity, sharing Quanta with a partner or a colleague can offset part of the next top-up without any change to how calls are billed.

The result is a system with few surprises: you know the price of the tokens, you know the charge for each call, and you know the rules for moving money in either direction.

04 · Published limits for heavy, sustained traffic

Quanta states its limits and enforces them, which is what allows a team to build on top of the service without discovering the boundaries during a busy period.

• › 120 requests per minute per key
• › 8 requests in flight per account
• › 25 MB per request body
• › 20 live keys per account

These figures suit the kind of work Quanta is designed for. Multiple keys let a team separate services, environments, or workloads. The per-key rate limit and the in-flight cap keep concurrency predictable, and the request body limit is known in advance rather than discovered by a failed upload.

Long-running streams are handled on the same published terms: there is no total-time cap on a stream, and a stream ends only after 300 seconds of upstream silence. For sustained generation jobs, that is a meaningful distinction, because a long answer is not treated as a timeout.

05 · One key and one balance across three protocols

Quanta supports the OpenAI-compatible /v1 protocol, Anthropic native, and Gemini native. The same key and the same balance work across all three, so a team can route different workloads without running separate billing relationships. The API base address is api.quantarelay.com, and the site is quantarelay.com.

Client pairing is specific, and Quanta publishes it plainly. Claude models work only from the official Claude Code and Claude Desktop clients, with Claude Code 2.1.285 or newer. Gemini models use the native protocol only, through Gemini CLI.

For teams using Cherry Studio, that means pairing it with GPT, DeepSeek, and the other non-Claude, non-Gemini models available on Quanta. Suitable options include GPT-6.1 Sol (gpt-6.1-sol) at $0.2083 / $1.01 / - / $0.0403, GPT-5.6 Sol (gpt-5.6-sol) at $0.4306 / $2.42 / - / $0.0347, DeepSeek V4 Flash (deepseek-v4-flash) at $0.2917 / $1.18 / - / $0.006944, Qwen3.7 Plus (qwen3.7-plus) at $0.1667 / $0.4861 / - / $0.0486, and models such as GLM-5.3, Kimi K3, Grok 4.7, and Hunyuan 4. The same set works with Codex, Cursor, Cline, OpenCode, OpenAI SDKs, and cURL.

That clarity is useful during setup. A team can choose the client that fits the work and the model that fits the budget, then watch the token breakdown for each request in the dashboard.

06 · Company payment, invoices, and where to ask

For a company account, payment is made by bank transfer from the company account. Invoices and advance quotes are arranged through the Contact page or the Messages page of the dashboard, so procurement and technical teams can work from the same place. Quanta operates a dashboard-based support channel and does not publish a support email address; the Contact page and dashboard messages are the routes to the team.

On data handling, the text users send is kept for 30 days for abuse handling and billing disputes, then deleted. Model answers are not stored. That retention window is stated plainly, which helps a team align its own record-keeping and review processes before sending production traffic.

Quanta combines a prepaid balance with no monthly fee, published token prices, per-call itemisation, defined limits, and straightforward company invoicing. For a team choosing an API relay, that is a short list of things to verify, and Quanta answers each of them directly. Open an account, apply the $0.50 trial credit, and measure your own first calls.