Claude Desktop API Relay Pricing: Quanta from $0.2083 per 1M

01 · Is one Quanta account enough for a heavy solo week?

Yes, for a power user who works alone but generates a heavy, sustained load, Quanta is built around the practical questions that show up in a busy week: how many calls can run at once, how fast requests can arrive, what each call actually costs, and how much money is left on the balance. Those answers are published rather than buried. You can read them before you start, and you can check every metered call in the dashboard afterwards.

Quanta is an AI API relay with a prepaid balance, no monthly fee and no minimum spend. There is no subscription structure to sign up for and no seat count to buy. You load the balance you want, use the API base address, and the meter follows real token consumption. For a solo operator, that means the bill tracks the work rather than the calendar.

Every call is metered on real tokens, and input, output, cache writes and cache reads are priced separately. The dashboard lists the tokens and the charge for every individual call, so when a heavy week produces a large total, you can trace it back to specific requests instead of guessing. The text you send is kept for 30 days for abuse handling and billing disputes and then deleted; model answers are not stored.

02 · What does pay-as-you-go pricing look like on Quanta?

Because Quanta charges by token category, the fastest way to understand the cost is to look at the model you plan to use. Prices are quoted per 1M tokens in approximate USD, and charges are made in CNY. A few examples from the current list make the shape of the pricing clear:

• › GPT-6.1 Sol is listed at $0.2083 input, $1.01 output, and $0.0403 cache read per 1M tokens.
• › Claude Haiku 4.5 is listed at $0.4167 input, $1.97 output, $0.4861 cache write and $0.0333 cache read per 1M tokens.
• › DeepSeek V4 Flash is listed at $0.2917 input, $1.18 output and $0.006944 cache read per 1M tokens.
• › Gemini 2.5 Flash is listed at $0.2361 input, $1.89 output, $0.2361 cache write and $0.0236 cache read per 1M tokens.

For monthly planning, the published worked examples use a 3:1 input-to-output ratio with no cache hits. On that example basis, Claude Sonnet 5.5 runs about $25.07 for 10 million tokens a month and about $250.69 for 100 million. Claude Opus 5.5 comes to roughly $34.58 for 10 million and $345.83 for 100 million. GPT-6.1 Sol is about $4.10 for 10 million tokens and $40.97 for 100 million, while Gemini 2.5 Flash is about $6.49 and $64.93 respectively. These are examples, not quotes, and your own mix of input, output and cache activity will produce its own figure on Quanta.

For a solo user, that visibility is the point. A heavy week is often a mix of long context runs and short iterative calls, and separate pricing for cache writes and cache reads means repeated context is not treated as if it were fresh input every time.

03 · Will heavy weeks hit rate limits or concurrency caps?

Quanta publishes and enforces its limits, so you can size your workflow against real numbers instead of discovering them mid-run:

• › 120 requests per minute per key for request rate.
• › 8 requests in flight per account for concurrency.
• › 25 MB per request body for payload size.
• › 20 live keys per account for key count.

The concurrency number is the one that matters most for a solo power user running parallel background jobs. Eight simultaneous requests is a defined ceiling rather than an opaque throttle, and because the limit is per account, a single heavy workload is metered consistently even if you have several keys open. If you want more separation between projects, the 20 live keys give you room to organise traffic without leaving Quanta.

Long-running streams are handled with a clear rule: there is no total-time cap on a stream, and it ends only after 300 seconds of upstream silence. That matters for agentic or batch-style work where a single response can legitimately take a long time. Quanta does not cut a stream short simply because it has been running for a while.

04 · How do Claude Desktop and other clients pair with Quanta?

Quanta speaks three protocols, and you use one key and one balance across all of them: OpenAI-compatible at /v1, Anthropic native and Gemini native. Pairing is specific to the client and the model family, so it is worth setting this out 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. Gemini models are not served over the OpenAI-compatible path.
• › GPT, DeepSeek and the other non-Claude, non-Gemini models pair with Codex, Cursor, Cherry Studio, Cline, OpenCode, OpenAI SDKs and cURL.

So if your heavy week alternates between a Claude Desktop session, a Gemini CLI run and a Codex or Cursor workflow, Quanta keeps one balance behind all of it rather than splitting your spend across separate accounts. The available model list includes Claude Fable 5, Claude Fable 5.1, Claude Haiku 4.5, Claude Opus 5, Claude Opus 5.5, Claude Sonnet 5, Claude Sonnet 5.5, DeepSeek V4 Flash, DeepSeek V4 Pro, Gemini 2.5 Flash, Gemini 3.1 Pro, Gemini 3.8 Flash, Gemini 4 Argon, Kimi K3, GPT-5.6 Sol, GPT-6 Astra, GPT-6.1 Sol, Qwen3.7 Plus, Qwen3.8 Max, Hunyuan 4, Grok 4.6, Grok 4.7 and GLM-5.3.

Switching between a fast model for routine steps and a heavier one for difficult passes is a normal pattern in a solo workflow, and on Quanta you can compare their real token charges in the dashboard once the run is done.

05 · How are top-ups, refunds and invoices handled?

Top-ups on Quanta are flexible and small to large: $1.39, $6.94, $13.89, $27.78, $69.44 and $138.89. Because there is no minimum spend, you can start with the smallest amount and add more in proportion to how heavy the week becomes. A $27.78 top-up, for instance, is also the point at which referral rewards can be withdrawn rather than moved into your balance.

The referral scheme rewards 19% of a referred account's top-ups and 5% one level further. If you bring in colleagues who also run heavy workloads, those rewards accumulate against the $27.78 withdrawal threshold, and you can either move them into your balance or withdraw them.

If a top-up is not what you expected, unused balance can be refunded within 15 days of the top-up. That short window is worth remembering, especially during a week when you are testing several models and the balance may sit partly unspent.

For business or expense handling, Quanta accepts bank transfer from a company account, and invoices and advance quotes are arranged through the Contact page or the Messages page of the dashboard. The same dashboard holds the per-call token and charge records that support your own reconciliation.

06 · How do you start on Quanta with little commitment?

Signing up adds $0.50 in trial credit, which is enough to make a first call and confirm that your chosen client pairs correctly with your chosen model before you commit real balance. From there, the sequence is simple: point your client at the Quanta API base address, use your key across whichever supported protocols you need, top up in the amount that matches your expected week, and read the per-call meter in the dashboard as the work accumulates.

For a solo power user, the appeal of Quanta is that nothing about the heavy week is hidden. Rates are listed, limits are stated, concurrency is defined, streams are not arbitrarily cut off, and every call shows its tokens and its charge. When the week is heavy, that clarity is what makes the spend understandable — and understanding the spend is what makes an account easy to keep open.