---
title: "Together AI Preemptible GPU compute preview; CLI v2.33.2 removes --scale-to-zero-window: 2026-09-10 | Harbyn"
description: "Together AI Preemptible GPU compute preview; CLI v2.33.2 removes --scale-to-zero-window, announced Sep 10, 2026."
canonical: "https://harbyn.com/changes/together/2026-09-10-preemptible-gpu-compute-and-cli-v2-33-2"
last-updated: 2026-09-29
---

# Together AI Preemptible GPU compute preview; CLI v2.33.2 removes --scale-to-zero-window

Markdown version of <https://harbyn.com/changes/together/2026-09-10-preemptible-gpu-compute-and-cli-v2-33-2>.

Breaking change - Together AI. medium severity. In effect.

Preemptible compute is in public preview for Together Kubernetes GPU clusters: set num\_preemptible\_gpus, and reclaimed nodes emit a TogetherPreemptionNotified event and give pods up to 300 seconds to exit. Together CLI v2.33.2 removes the --scale-to-zero-window flag from tg beta endpoints deploy and tg beta endpoints update.

## Facts

| Fact | Value |
| --- | --- |
| Vendor | [Together AI](https://harbyn.com/vendors/together.md) |
| Kind | Breaking change |
| Severity | medium |
| Status | In effect |
| Announced | Sep 10, 2026 |
| Added to Harbyn | Sep 29, 2026 |
| Lead time | No dated deadline |
| Vendor page | <https://docs.together.ai/docs/changelog> |
| Review | Reviewed by a person, extracted by a deterministic parser |

## Find it in your code

```sh
npx harbyn scan
```

This change names no identifier the scanner can match line by line, so the scan tells you whether your repository uses Together AI, from its packages and API hosts. Nothing leaves your machine.

## Sources

Where this event comes from: the page we read, when we read it, and the hash of that snapshot.

- <https://docs.together.ai/docs/changelog/rss.xml> (changelog, fetched 2026-09-29 02:31 UTC, sha256 `10415e14c75c8a486ad7c21bb4c80df09f1a4b5f3fa031a1c9040f4ea4657ce3`)

## Related changes

### More from Together AI

- [Together CLI v2.33.2 improvements](https://harbyn.com/changes/together/2026-09-10-preemptible-compute-for-gpu-clusters-preemptible-compute-is-now-in-public-pr-2.md) (Behavior change, announced Sep 10, 2026)
- [four serverless models removed on September 14, 2026](https://harbyn.com/changes/together/2026-09-01-four-serverless-models-removed-september-14-2026.md) (Retirement, effective Sep 14, 2026)
- [H100 80GB dedicated endpoint price cut to $3.99 per hour](https://harbyn.com/changes/together/2026-09-01-h100-dedicated-endpoint-price-cut.md) (Pricing, announced Sep 1, 2026)
- [lower serverless pricing for Qwen3.7-Max and Qwen3.8-Flash](https://harbyn.com/changes/together/2026-09-22-qwen-serverless-price-cut.md) (Pricing, effective Sep 22, 2026)

### Changes at other vendors

- [Amazon Selling Partner API: October listing attribute usage and enumeration value updates](https://harbyn.com/changes/amazon-sp-api/2026-10-14-sp-api-updates-new-listings-items-api-ivp-values-in-eu5-jp-and-ca-stores-oct-2.md) (effective Oct 26, 2026)
- [Supabase: four CBC-mode TLS 1.2 cipher suites will be removed from project HTTP APIs](https://harbyn.com/changes/supabase/2026-10-12-an-outdated-subset-of-tls-1-2-ciphers-will-be-disabled-on-project-http-apis.md) (effective Nov 10, 2026)
- [Google Workspace APIs: OAuth clients must be marked as agentic to call Google Workspace MCP servers](https://harbyn.com/changes/google-workspace/2026-10-12-oauth-clients-must-be-marked-as-agentic-to-call-google-workspace-mcp-servers.md) (effective Nov 11, 2026)
- [Bitbucket Cloud API: Git Credential Manager upgrade required for OAuth 2.0 Git authentication](https://harbyn.com/changes/bitbucket/2026-10-13-git-credential-manager-upgrade-required-for-oauth-2-0-git-authentication.md) (effective Nov 30, 2026)

[All upcoming deadlines](https://harbyn.com/deadlines.md)