- IXP release notes
- IXP in Flow release notes
- Communications Mining release notes
- Unstructured and complex documents release notes
October 2026
October 2026 release notes for Unstructured and Complex Documents in Automation Cloud.
For IXP service-level updates, refer to the IXP release notes.
For other IXP capability updates, refer to the Communications Mining release notes.
October 30, 2026
New model picker
The extraction model picker on the Build tab has been redesigned to help you choose the right model for your project:
- Models are grouped into categories: Available, Preview, and Custom, for the models you bring through your own LLM subscriptions.
- When a model does not respect your tenant's region, the picker shows the region where requests are routed.
- You can filter models by category and by provider.
- Models scheduled for deprecation are marked in the picker, so you can switch to another model in advance.
For details, check Model configuration.
Claude models in preview
The following Anthropic Claude extraction models are now available in preview:
- Claude Sonnet 4.6
- Claude Sonnet 5
Claude models have higher latency than other extraction models, so predictions can take longer, especially for long documents. For details on what requests are routed per region, check AI features and model routing.
Deprecating Gemini 3.6 Flash and Gemini 3.7 Flash
The Gemini 3.6 Flash and Gemini 3.7 Flash extraction models, previously available in preview, are deprecated in favor of Gemini 3.8 Flash. If your projects use Gemini 3.6 Flash or Gemini 3.7 Flash, switch them to Gemini 3.8 Flash. For details, check Switching from one model to another.
Bug fixes
- Prediction requests sometimes stalled for several minutes when multiple requests loaded the same model version at the same time.
October 23, 2026
GPT-5.1, GPT-5.4, and Gemini 3.5 Flash Lite are generally available
The following extraction models are now generally available, and no longer marked as preview:
- gpt-5.1-2025-11-13
- gpt-5.4-2026-03-05
- Gemini 3.5 Flash Lite
You can select these models from the Model configuration option on the Build tab. For details, check Model configuration. You can also run them on your own LLM subscriptions. For details, check Bring your own subscription (BYOS).
October 14, 2026
Bug fixes
Predictions made with Gemini 2.5 Flash sometimes returned a confidence value of 1.0 for all predicted values, because the model did not always return log probabilities.
October 6, 2026
Introducing Gemini 3.5 Flash
The Gemini 3.5 Flash extraction model is now generally available for a set of customers, with broader access to follow. For all other customers, the model remains available in preview. For details on what requests are routed per region, check AI features and model routing.