Starting the night of Sunday, September 27, 2026, Tulip AI Agents will move off Claude Sonnet 4, which is being retired by our model provider. Depending on your deployment region, you’ll move to Claude Sonnet 5 (where available) or Claude Sonnet 4.6 (all other regions). This applies across both weekly and LTS releases. No action is required on your part; this is a backend configuration change, not a release upgrade.
Summary of changes
| Before | After (from Sept 27, 2026) | |
|---|---|---|
| Model powering AI Agents | Claude Sonnet 4 | Claude Sonnet 5 (available regions) or Claude Sonnet 4.6 (all other regions) |
| Applies to | — | All regions, weekly and LTS, simultaneously |
| Action required | — | None |
What’s changing
1. The model behind Agent Builder and published agents: Every agent will run on the new model starting Sept 27.
2. Which model you get depends on region: Sonnet 5 isn’t yet available in every region where Tulip customers are hosted. Where it is, your Agents move to Sonnet 5; everywhere else, your Agents move to Sonnet 4.6. Both are meaningful upgrades over Sonnet 4 in our testing.
3. Why this isn’t purely routine: For most Tulip functionality, we validate the integration with a model provider rather than the model’s behavior, and swaps happen invisibly. Agents are different: you write and own your agents’ prompts, so a model change can shift how an agent responds even though nothing in your app changed. That’s the reason for this notice.
When does this take effect?
Overnight on Sunday, September 27, 2026, across all customer regions. There is no separate phased rollout beyond the region-based model assignment above.
What stays the same?
- Your agents, prompts, and configurations are untouched — only the underlying model changes.
- No release upgrade or action is required to receive this update.
- You can see which model your agent is currently running via in-app model alias display.
Frequently asked questions
How do I know if my region is getting Sonnet 5 or Sonnet 4.6? In-app model alias display
Will my agents behave differently after this change? Possibly, for agents with tightly-tuned or very specific prompts. We ran a structured comparison across Sonnet 4, 4.6, and 5 on representative scenarios and found both replacement models performed as well as or better than Sonnet 4. We recommend spot-checking production agents after the change.
Do I need to update anything? No, this is a backend configuration change only.
Will there be any downtime assciated with this change? Nope, this change will take place immedately with no downtime.
Staying informed
- Reach out via support@tulip.co or your CSM with questions.