Skip to content
Back to feed

November 15, 2026 · Fireworks AIDeprecation

Action needed

Fireworks dataset creation and upload deprecated, now returns 501 Not Implemented

Fireworks no longer stores new datasets. Creating, updating, uploading, validating, previewing, splitting, and downloading datasets now return 501 Not Implemented across the REST/gRPC API, firectl, and the Python SDK.

  • What remains available: List Datasets, Get Dataset, and Delete Dataset (firectl dataset list, firectl dataset get, and firectl dataset delete) remain active so you can inspect metadata for and delete existing dataset records.
  • Training: Load and tokenize training data directly from local files or your own cloud storage with the Training API — no dataset resource is created or stored on Fireworks. See Migrate from Managed Training.
  • Batch workloads: Run bulk asynchronous workloads from local JSONL files on preemptible deployments. See Migrate from Batch API.

All Fireworks AI platform changes