# Together opens custom model training service with dedicated GPU allocations

_Published Monday, October 19, 2026 at 2:19 PM EDT · AI · Latest · Tier 2 — Notable_

![Together opens custom model training service with dedicated GPU allocations — Primary](https://cdn.prod.website-files.com/69654e88dce9154b5f12070c/6ad63420bade9a234b1c0595_6ad6340405d0cf88b81a11a2_og-image-introducing-together-post-training.png)

Together says its Post-Training service is now available to everyone after months of early access, allowing developers to define their own training loops for reinforcement learning and supervised training. The service allocates dedicated GPUs to each customer's model resources and supports multiple lightweight model adapters sharing the same GPU capacity.

Customers can supply their own data, environments and training objectives. Together bills resources per GPU-minute and offers B200 usage at $9 per GPU-hour for the first eight weeks. Developers can update all model weights instead of using adapters, but that option is limited to select models.

## Sources

- [Together.ai](https://www.together.ai/blog/together-post-training)

---
Canonical: https://techandbusiness.org/newswire/9l8Qgo85nDTqmbiZf25QwO
Published: 2026-10-19T18:19:19.207Z
Story chronology: 2026-10-19T12:00:00.000Z
Retrieved: 2026-10-19T21:26:39.989Z
Publisher: Tech & Business (techandbusiness.org)