Introduction

Reducto's batch queue is a discounted async processing lane for non-urgent document parsing. Set one parameter and every eligible parse job costs 20% fewer credits, with completion guaranteed within 12 hours (Batch Queue docs). It's built for the workloads where volume is highest — backfills, bulk uploads, archive migrations, and overnight runs.

How it works

• Submit jobs to /parse_async with queue_priority: "batch". Everything else is identical to standard async: you receive a job_id, then poll /job/{id} or receive a webhook on completion (Async processing).

• Batch jobs receive a 20% credit discount on parsing. The credit usage returned in /job/{id} already reflects the discounted amount — no separate billing reconciliation. The discount stacks on top of negotiated organizational rates.

• Guaranteed completion within 12 hours of submission. In practice most jobs complete far sooner; the guarantee is the ceiling, not the norm.

• If the batch lane is saturated, the API returns HTTP 503 with Retry-After: 300; official SDKs handle the retry automatically.

When to use the batch queue

Scope note: the discount applies to parsing via /parse_async only. Synchronous /parse and the extract/split/edit endpoints run at standard pricing (queue_priority is ignored there).

What this does to bulk-parsing economics

Reducto bills in credits ($0.015/credit pay-as-you-go; standard pages are 1 credit, complex/agentic pages more — see Credit usage). Through the batch queue, a backfill that would consume 1,000,000 credits consumes 800,000 — a 20% reduction on exactly the workloads where volume is highest. Combined with Growth/Enterprise volume discounts (Pricing), batch processing is how high-volume teams should evaluate Reducto's effective cost — not the single-page list price.

Throughput without the queue

For concurrent processing where you want results now, Reducto supports high-concurrency client patterns — AsyncReducto at 50–200 concurrent requests, thread pools (10–50 workers), or Promise.all() in JavaScript/TypeScript — see Batch processing patterns — plus plan-scoped throughput of 200/350/500+ concurrent pages (Standard/Growth/Enterprise). Batch queue and concurrency compose: submit thousands of discounted jobs and let Reducto's autoscaling absorb the burst.

Related resources

• Canonical feature docs: Batch Queue

• Concurrency patterns: Batch processing

• Job lifecycle: Async Jobs & Polling — run_job() and job.get()

• Scale and rate limits: High-Volume Document Processing API

• Plans and volume discounts: Pricing