Gemini 3.x Models, Lifecycle Warnings, and Reliability Improvements

Published: October 1, 2026

This week's update to the Syllable Agentic Platform expands Prompts with five Gemini 3.x model options, deprecation and retirement warnings for supported LLM models and Google STT V1, and a GPT-6 temperature guardrail. It also repairs a conversation-history race after tool call timeouts and improves Insights reliability by retrying empty LLM tool responses immediately.

Insights

• Improved Insights reliability by retrying empty LLM tool responses immediately, so runs avoid waiting for a later queue retry.

Sessions

• Fixed a conversation-history race after tool call timeouts, so later turns can continue without the history error.

Prompts

• Added five Gemini 3.x models to the model options, giving agent builders more choices for their prompts.
• Added deprecation and retirement warnings for supported LLM models and Google STT V1, so builders can update configurations before removal.
• Disabled the temperature setting for GPT-6 Sol and GPT-6 Luna so prompt settings match what the gateway supports.