---
title: "Submit"
method: POST
path: "/api/inference"
tags: ["Inference"]
---

# Submit

`POST /api/inference`

Submit a serverless inference request.

Schedules a short-lived GPU job that runs the specified model on the
provided inputs. Returns a job_id to poll for results.

## Request body

- InferenceRequest
  - `model` string, required — Model name or HuggingFace repo (e.g. 'distilbert-base-uncased-finetuned-sst-2-english')
  - `inputs` union, required — Text input(s) for inference
    - string[]
    - string
  - `gpu_model` string — Preferred GPU model or 'any'
  - `max_tokens` integer
  - `temperature` number
  - `timeout_sec` integer

## Response `200`

Successful Response

- unknown

## Other responses

- `422` — Validation Error

## Changes

- **2026-06-08** `cbe3ad94235c` — 1 info
  - endpoint added
- **2026-04-11** `f1121e9c701c` — 1 breaking
  - api path removed without deprecation
- **2026-04-05** `f2668775365c` — 1 info
  - endpoint added
- **2026-04-03** `aca6ce6b2785` — 1 breaking
  - api path removed without deprecation

[Change history](https://skmtc.dev/aabiro/apis/xcelsior/changes/api/inference/post.md)

---

[API](https://skmtc.dev/aabiro/apis/xcelsior.md) · [All operations](https://skmtc.dev/aabiro/apis/xcelsior/llms.txt) · [OpenAPI document](https://skmtc-service-production.skmtc.workers.dev/v1/apis/aabiro/xcelsior/revisions/32211b6f9d65/schema)