Message Batches
Submit many message requests as one batch and pull results when it ends
Message Batches let you submit a set of message requests as a single job, then retrieve the results when processing finishes. The shape follows the Anthropic Message Batches API.
Batches run on the Anthropic-compatible surface at /v1/messages/batches. The OpenAI-style /v1/batches endpoint is not offered, so use the Anthropic batch endpoints below.
Create a batch
POST /v1/messages/batches takes a requests array. Each entry has a custom_id you choose and a params object holding the same body you would send to /v1/messages.
curl https://api.ablatic.ai/v1/messages/batches \
-H "x-api-key: $ABLATIC_API_KEY" \
-H "anthropic-version: 2023-06-01" \
-H "Content-Type: application/json" \
-d '{
"requests": [
{
"custom_id": "req-1",
"params": {
"model": "talos",
"max_tokens": 1024,
"messages": [{ "role": "user", "content": "Summarize Q1." }]
}
},
{
"custom_id": "req-2",
"params": {
"model": "talos",
"max_tokens": 1024,
"messages": [{ "role": "user", "content": "Summarize Q2." }]
}
}
]
}'Send an Idempotency-Key header to make a create call safe to retry.
Manage a batch
GET /v1/messages/batches/{id} retrieve one batch
GET /v1/messages/batches list batches
POST /v1/messages/batches/{id}/cancel cancel a batch
DELETE /v1/messages/batches/{id} delete a batch
GET /v1/messages/batches/{id}/results stream resultsRead results
When a batch reaches processing_status of "ended", fetch its results endpoint. The body streams as application/x-ndjson, one JSON result per line.
Limits
- Per batch: up to 100000 requests and 256 MiB.
- Per user: up to 50 batches in flight at once.
- Results stay available for 29 days.
Each entry inside params is a standard Messages request, so max_tokens is required there. See the Anthropic compatibility page for the full body.