Thinking and Tool Blocks

Some AI models show their reasoning process before giving a final answer. When tools are in use, the model's tool calls and results are also shown. OptimaChat displays these as collapsible blocks inside the conversation.

Thinking blocks

A thinking block shows the model's internal reasoning — the step-by-step thought process the model worked through before writing its response. Not all models produce thinking blocks; this depends on the model and whether extended thinking is enabled.

Thinking blocks are collapsed by default once the model has finished responding. You can click a thinking block to expand and read it in full.

Tool blocks

A tool block (also called an agent run block) summarises the tool calls a model made during a response. See Using Tools for a full explanation of how the model uses tools and what the status labels mean.

Tool blocks are collapsed by default when the run completes. Click one to expand it and see the individual tool calls, their inputs, and their outputs.

Keeping blocks expanded

By default, thinking blocks and tool blocks collapse automatically when the model finishes a run. You can change this behaviour using the toggles in the settings panel:

• Thinking Blocks — when enabled, thinking blocks stay expanded after the model finishes responding
• Tool Blocks — when enabled, tool (agent run) blocks stay expanded after the run completes

These settings apply to all new messages in the current session. They do not retroactively expand blocks from earlier in the conversation.

Why use these blocks

• Thinking blocks help you understand how the model arrived at an answer, which is useful for verifying reasoning or diagnosing unexpected responses.
• Tool blocks give you a full audit trail of what the model searched for, what it retrieved, and how it used that information — useful for trust and transparency.

For most everyday use, keeping both collapsed is the tidiest experience. Expand them when you want to inspect the model's work in detail.