Route large volume AI jobs to lowest-cost eligible provider
Run a specific model or a curated workflow product on the cheapest eligible provider lane. Pay for actual token usage — with a pre-dispatch cost preview and a settlement receipt.
- No provider lock-in
- OpenAI Batch compatible
- Webhook delivery
Route preview
- 1One batch fileSubmit
Upload JSONL or start from a workflow product.
- 2Lowest eligible routeEstimate
Versioned and auditable before dispatch.
- 3Artifacts & receiptReceive
Artifact download, webhook, or storage target.
How it works
Four steps from submission to delivered artifacts.
Submit a batch
Pick a direct route or a curated workflow contract. Set splitting policy and delivery target — every batch ships with our 24-hour SLA.
Preview the lowest route
Batchrouter evaluates eligible provider lanes and shows the cheapest route with a pre-dispatch cost preview and a routing explanation.
We route & retry
Dispatched to the eligible lane, or split into internal lanes while you keep one customer batch.
Receive artifacts
Finished outputs and a settlement receipt land via artifact download, webhook, or your storage target.
Featured workflows
Start from a curated outcome contract instead of choosing a model first.
Document classifier
Tag PDFs and emails into your taxonomy with confidence scores and human-review triggers.
- SLA
- 24h
- Output
- Schema-validated
Structured extractor
Pull typed fields out of invoices, contracts, and receipts against your JSON schema.
- SLA
- 24h
- Output
- Schema-validated
Rerank & score
Score candidate documents or chunks for relevance and ship results to your retrieval pipeline.
- SLA
- 24h
- Output
- Schema-validated
One POST, cheapest lane.
Submit a JSONL batch and get the cheapest eligible lane with a pre-dispatch cost preview and routing rationale. Accept to dispatch the lane.
quote_idStable handle for accept / dispatch / settle.laneCheapest eligible provider lane, with routing rationale.estimated_totalIllustrative pre-dispatch cost preview — you pay for actual token usage.sla_hoursDelivery window for the batch.
$ curl https://api.batchrouter.com/v1/quotes/model \
-H "Authorization: Bearer bk_live_..." \
-H "Content-Type: application/json" \
-d '{"models":["claude-3-haiku"],"task":"responses","estimated_items":4218}'
{
"quote_id": "qt_8K2xR9pL",
"lane": "anthropic.claude-3-haiku.batch",
"items": 4218,
"estimate_until": "2026-05-24T18:42:00Z",
"estimated_total": { "amount": "12.47", "currency": "USD" },
"sla_hours": 24
}Illustrative example — values vary per batch. You pay for actual token usage at completion.
A clean HTTP API for agents to run the full batch lifecycle
No dashboard required. Sign up, estimate, dispatch, retrieve — all from code.
HTTP signup
Register an org and get an API key in a single POST. No dashboard required.
Versioned estimates
Returned quote_id values snapshot the route and settle against actual token usage.
Webhook delivery
Results push to your endpoint, with artifact download and storage targets as alternatives.
OpenAI-compatible
Drop in existing batch JSONL files from the OpenAI Batch API — no conversion.
Ship one batch. See the savings.
Preview in seconds. Pay for actual token usage on the lane that wins.