Save on document processing with batch inference

Run extraction, classification, and labeling as a single job: one line per document, one file, one submission. Request JSON output, collect results within 24 hours, and pay less than OpenRouter batch inference.

Request accessRead the docs
API example

Request JSON on every line

Submit the job with HTTP or the OpenAI SDK. Include a JSON schema on each chat body so every document returns as structured data you can load without cleanup.

See structured outputs 
curl https://us.api.lararouter.com/v1/batches \
  -H "Authorization: Bearer $LARAROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "input_file_id": "file_9Hs4TnQ2",
    "endpoint": "/v1/chat/completions",
    "completion_window": "24h"
  }'
How it works

One file in, one file out

Everything runs through the same jobs API: upload a JSONL file, create a batch, and download the results. There is no interactive tier with a premium price.

Structured JSON from every document

Each line is a chat request with a JSON schema, so results arrive ready to store—not free-form text you have to clean up later.

Retry only the failures

Failed lines land in the error file with the custom_id you assigned. Rebuild a file from those ids and resubmit without reprocessing the rest.

Stronger models at batch pricing

Long documents often need a larger model. Batch pricing makes high-quality extraction practical at volumes that do not justify interactive rates.

Models

Models for document processing

Recommended models for extraction, classification, and labeling, with the batch rate each one bills.

BatchPer 1M tokens
Recommended for
CohereCommand R 35BSummarizationDocument Processing
US$ 0,25
23% below OpenRouter batch
US$ 0,75
21% below OpenRouter batch
DeepSeekDeepSeek-R1-Distill-Qwen 14BSummarizationDocument Processing
US$ 0,15
21% below OpenRouter batch
US$ 0,15
21% below OpenRouter batch
DeepSeekDeepSeek-R1-Distill-Llama 70BSummarizationDocument Processing
US$ 0,40
20% below OpenRouter batch
US$ 0,40
20% below OpenRouter batch
MetaLlama 3.3 70BSummarizationDocument Processing
US$ 0,44
20% below OpenRouter batch
US$ 0,44
20% below OpenRouter batch
MetaLlama 3.1 70BSummarizationDocument Processing
US$ 0,44
20% below OpenRouter batch
US$ 0,44
20% below OpenRouter batch
MistralMistral Small 3 24BSummarizationDocument Processing
US$ 0,30
20% below OpenRouter batch
US$ 0,30
20% below OpenRouter batch
MistralMixtral 8×7BSummarizationDocument Processing
US$ 0,30
20% below OpenRouter batch
US$ 0,30
20% below OpenRouter batch
MistralCodestral 22BDocument Processing
US$ 0,30
20% below OpenRouter batch
US$ 0,30
20% below OpenRouter batch
QwenQwen2.5 14BSummarizationDocument Processing
US$ 0,10
26% below OpenRouter batch
US$ 0,40
20% below OpenRouter batch
QwenQwen2.5-Coder 7BDocument Processing
US$ 0,10
20% below OpenRouter batch
US$ 0,15
25% below OpenRouter batch
QwenQwen2.5-Coder 14BDocument Processing
US$ 0,10
26% below OpenRouter batch
US$ 0,40
20% below OpenRouter batch
QwenQwen2.5 32BSummarizationDocument Processing
US$ 0,40
20% below OpenRouter batch
US$ 0,40
20% below OpenRouter batch
QwenQwen2.5 72BSummarizationDocument Processing
US$ 0,45
22% below OpenRouter batch
US$ 0,60
20% below OpenRouter batch
QwenQwen2.5-Coder 32BDocument Processing
US$ 0,40
20% below OpenRouter batch
US$ 0,40
20% below OpenRouter batch
Savings are measured against OpenRouter batch inference for the same model.

Start processing documents

Upload a file, create a batch, and download structured results within 24 hours. Enterprise plans deliver in as little as 4 hours.

Request accessCompare model pricing