OpenCode pay-as-you-go pricing: Quanta usage records, $0.50 credit

01 · What Quanta's usage records show you

Teams running heavy, sustained workloads do not need another rounded-up summary of their traffic. They need a record they can trust line by line. Quanta's usage records are built for that. Every call is metered on real tokens, and the dashboard lists the tokens and the charge of every call. Input, output, cache writes and cache reads are priced separately, so each entry shows what was sent, what came back, what was written to cache and what was read back from it.

That granularity matters when a job runs continuously rather than occasionally. A single monthly total can hide an expensive pattern: repeated context, long outputs, or a cache written far more often than it is read. Quanta keeps those details visible instead of averaging them away, which is the difference between a billing total you have to interpret and an operations record you can act on.

Quanta is pay-as-you-go only. The account runs on a prepaid balance, with no monthly fee and no minimum spend. The usage records are therefore the natural centre of the account rather than a reporting layer attached to a subscription.

02 · Reading tokens, cache activity and the charge on each call

Because input, output, cache write and cache read are charged on their own terms, the useful question is not simply "how much did we spend?" but "which part of the traffic produced the charge?" Quanta's per-call records let you answer that directly, and the price list behind them is public rather than hidden behind a quote.

• › GPT-6.1 Sol is listed at $0.2083 for input and $1.01 for output per 1M tokens, with cache reads at $0.0403.
• › Qwen3.7 Plus is listed at $0.1667 for input and $0.4861 for output per 1M tokens, with cache reads at $0.0486.
• › DeepSeek V4 Flash is listed at $0.2917 for input and $1.18 for output per 1M tokens, with cache reads at $0.006944.
• › Claude Fable 5.1 is listed at $4.38 for input, $21.88 for output, $5.39 for cache writes and $0.2222 for cache reads per 1M tokens.

All quoted prices are approximate USD figures; charges are made in CNY. Two worked examples show how the metering behaves at volume. As an example, at a 3:1 input-to-output ratio with no cache hits, Claude Opus 5.5 at 10 million tokens a month comes to approximately $34.58, and 100 million to approximately $345.83. As another example, GPT-6.1 Sol at the same shape comes to approximately $4.10 for 10 million tokens and approximately $40.97 for 100 million. These are examples, not quotes, and your own records will reflect the traffic you actually send.

03 · Concurrency, rate limits and the shape of sustained traffic

Heavy workloads fail in predictable ways: too many requests at once, a body that is larger than the service accepts, or a key that is asked for more per minute than the account is allowed. Quanta publishes those limits and enforces them, so capacity planning starts with stated numbers instead of guesswork.

• › 120 requests per minute per key.
• › 8 requests in flight per account.
• › 25 MB per request body.
• › 20 live keys per account.
• › Streams carry no total-time cap and end only after 300 seconds of upstream silence.

Those figures sit alongside the per-call records. When traffic is running hot, you can look at what is being metered, see how many calls are in flight, and check the charge attached to each request as it lands. For a team that runs scheduled pipelines, batch jobs or long-running agents, that combination is the practical part of operations: published ceilings, enforced behaviour and a ledger that shows the cost of the work rather than an estimate of it.

04 · OpenCode setup: one key and one balance across three protocols

Quanta exposes an OpenAI-compatible (/v1) path, an Anthropic native path and a Gemini native path, and one key and one balance work across all three. That is convenient for teams running several tools against the same budget, because usage from different clients still lands in the same Quanta records.

Client pairing is specific, and it is worth setting up carefully. OpenCode pairs with GPT, DeepSeek and the other non-Claude, non-Gemini models on the Quanta list, including Qwen, Kimi, Hunyuan, Grok and GLM. Claude models are a separate path: they work only in 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 an OpenCode team, that means you can point the client at Quanta's OpenAI-compatible endpoint, use a single key, and then read back what each request cost. Start with a $0.50 trial credit on sign-up, top up as the records show you real usage, and adjust model selection based on what the per-call entries actually charge.

05 · Top-ups, company invoices and refunds

Because Quanta has no monthly fee and no minimum spend, top-ups are the only commitment you make. Available top-up amounts are $1.39, $6.94, $13.89, $27.78, $69.44 and $138.89, so a team can size a top-up to the traffic it has already measured in the dashboard instead of buying capacity it may not use.

• › Unused balance can be refunded within 15 days of the top-up.
• › Referral rewards are 19% of a referred account's top-ups and 5% one level further; rewards can be withdrawn from $27.78 or moved into the balance.
• › Company payment is made by bank transfer from a company account, and invoices and advance quotes are arranged through the Contact page or the Messages page of the dashboard.

For finance teams, that last point is the one that removes friction. Invoices and advance quotes are handled in the same place as the operational messages, so a purchase order conversation does not have to happen somewhere else while the engineering team watches usage records. No monthly fee and no minimum spend mean the invoice follows the work.

06 · What Quanta stores, and for how long

Usage records and data handling are separate questions, and Quanta answers both. The text you send is kept for 30 days for abuse handling and billing disputes, and then deleted. Model answers are not stored.

That distinction is worth noting when you read the dashboard. The records are there to show tokens and charges, and they support the operational questions a busy team asks each week: which model is carrying the cost, whether cache reads are doing the work they should, and whether the traffic pattern matches the top-up you bought. Open an account at Quanta, fund it with the $0.50 trial credit, and let the first week of records tell you what the next month should cost.