Qwen3.8 Max API Relay Pricing: Quanta's Nightly Batch Month

01 · What a month of nightly batch jobs costs on the Quanta rate card

Nightly batch work is easy to underestimate. A single evening run may look small on a chart, but across a month the token counts accumulate into a line item that a team has to defend. Quanta handles this by publishing a rate card with a price for every metered category, so the estimate you build before the first run is the same shape as the invoice you review afterwards.

Quanta quotes prices per 1M tokens and separates input, output, cache writes and cache reads. For example, Qwen3.8 Max is listed at $0.8611 per 1M input tokens, $2.58 per 1M output tokens and $0.1028 per 1M cache read tokens. Claude Sonnet 5.5 is listed at $1.25 input, $6.28 output, $1.57 cache write and $0.0917 cache read. Gemini 2.5 Flash is listed at $0.2361 input, $1.89 output, $0.2361 cache write and $0.0236 cache read. There is no bundled allowance and no monthly fee to negotiate around; Quanta is pay-as-you-go, and the balance you hold is the balance you spend.

• › Input tokens are billed on their own line, separate from output.
• › Cache writes and cache reads carry their own rates, so repeated context has a visible cost.
• › The dashboard lists the tokens and charge of every call, which makes a month-end reconciliation a matter of reading figures rather than interpreting a summary.

02 · Published limits for sustained nightly throughput

A batch schedule only works if the relay tells you what it will accept. Quanta publishes and enforces its limits rather than leaving them to be discovered under load: 120 requests per minute per key, 8 requests in flight per account, 25 MB per request body and 20 live keys per account.

Those numbers shape how a nightly job should be written. Eight requests in flight per account means concurrency is a scheduling decision you make in your own runner, not a queue that quietly builds somewhere else. Twenty live keys give a team room to separate workloads — one key for nightly batch, another for interactive traffic — while the per-key rate of 120 requests per minute keeps a burst from monopolising the account.

Long-running streams are also described plainly. Streams have no total-time cap and only end after 300 seconds of upstream silence. For a batch pipeline that streams partial output back to your storage, that is a specific, checkable condition rather than a vague promise about generous timeouts.

03 · Choosing models for a month of nightly runs

The Quanta rate card covers Claude, Gemini, GPT, DeepSeek, Qwen, Kimi, Hunyuan, Grok and GLM models. Which one fits depends on the shape of the job and on the client you already run.

Four worked examples from the rate card, each using a 3:1 input-to-output mix with no cache hits:

• › Claude Opus 5.5 at 10 million tokens a month is approximately $34.58, and 100 million tokens is approximately $345.83.
• › Claude Sonnet 5.5 at 10 million tokens a month is approximately $25.07, and 100 million tokens is approximately $250.69.
• › GPT-6.1 Sol at 10 million tokens a month is approximately $4.10, and 100 million tokens is approximately $40.97.
• › Gemini 2.5 Flash at 10 million tokens a month is approximately $6.49, and 100 million tokens is approximately $64.93.

These are examples, not quotes, but they show the arithmetic a team can repeat: take your monthly input and output volume, apply the two rates for the model you intend to use, and compare. For very high-volume classification or extraction work, the lower end of the card is worth measuring. Qwen3.7 Plus, for instance, is listed at $0.1667 input and $0.4861 output. For heavier reasoning on the same card, Qwen3.8 Max sits at $0.8611 input and $2.58 output, while DeepSeek V4 Flash is listed at $0.2917 input and $1.18 output.

Because cache reads are priced separately and considerably lower than fresh input on most entries, a nightly job that reuses a stable context can be modelled with the cache read rate rather than the input rate. Quanta does not hide that lever behind a plan; it puts the four prices side by side.

04 · How Quanta meters and itemises the work

Billing on Quanta is prepaid. You add funds to a balance, and each call is metered on real tokens. There is no monthly fee and no minimum spend, which suits a workload whose volume moves from week to week: a quiet month simply consumes less of the balance you have already funded.

Two details matter for a team running unattended jobs. First, the dashboard lists the tokens and the charge of every individual call, so when a nightly run costs more than expected, you can identify which run and which request produced the difference. Second, retention is stated directly: the text users send is kept for 30 days for abuse handling and billing disputes, then deleted, and model answers are not stored. For a pipeline that sends customer records through a model, that is a fact worth having in writing next to the runbook.

Top-ups are available at $1.39, $6.94, $13.89, $27.78, $69.44 and $138.89, and unused balance can be refunded within 15 days of the top-up. New accounts receive $0.50 in trial credit on sign-up, enough to make a first measured call before committing to a larger amount.

05 · Paying for Quanta as a company

Quanta is built for teams, and company payment runs through a bank transfer from a company account. Invoices and advance quotes are arranged through the Contact page or the Messages page of the dashboard, so procurement can request the paperwork it needs without a separate sales cycle. There is no support email address to hunt for; the two dashboard routes are the way in.

Referral rewards are also part of the balance rather than a separate scheme: 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, which lets a team use them for the next month of batch runs instead of holding them idle.

For a finance lead, the useful summary is short. Prepaid balance, no monthly fee, no minimum spend, itemised per-call charges, published limits, and invoices handled through the Contact page or the Messages page. Nothing about the pricing changes shape between the sales conversation and the dashboard.

06 · Pairing clients and starting a Quanta account

Quanta serves three protocols — OpenAI-compatible at /v1, Anthropic native and Gemini native — and one key and one balance work across all three. The pairing rules are specific, so set them up correctly from the start:

• › Claude models work only from the official Claude Code and Claude Desktop clients, using Claude Code 2.1.285 or newer.
• › Gemini models use the native protocol only, through Gemini CLI.
• › Codex, Cursor, Cherry Studio, Cline, OpenCode, OpenAI SDKs and cURL pair only with GPT, DeepSeek and the other non-Claude, non-Gemini models on the Quanta card, including Qwen3.8 Max, Qwen3.7 Plus, Kimi K3, Hunyuan 4, Grok 4.6, Grok 4.7 and GLM-5.3.

A sensible first month looks like this. Open a Quanta account, claim the $0.50 trial credit, and run a representative slice of the nightly workload against the model your client supports. Read the per-call token and charge lines in the dashboard, compare them with your own 3:1 estimate, and then decide which model and which cache pattern to standardise on. Top up at the level that suits the volume you measured, and arrange the invoice through the Contact page if your company pays by bank transfer.

What makes Quanta worth a look for heavy, sustained workloads is the plainness of it: published per-token rates, published limits, one balance across protocols, and a dashboard that shows the tokens and charge for every call. For a month of nightly batch jobs, that is enough to build a budget, and enough to check it afterwards.