Google Gemini now supports thinking tokens

Here is a sample as a get request

#!/bin/bash
set -e -E

GEMINI_API_KEY="$GEMINI_API_KEY"
MODEL_ID="gemini-3-flash-preview"
GENERATE_CONTENT_API="streamGenerateContent"

cat << EOF > request.json
{
"contents": [
{
"role": "user",
"parts": [
{
"text": "INSERT_INPUT_HERE"
},
]
},
],
"generationConfig": {
"temperature": 0,
"thinkingConfig": {
"thinkingLevel": "MINIMAL",
},
},
}
EOF

curl \
-X POST \
-H "Content-Type: application/json" \
"https://generativelanguage.googleapis.com/v1beta/models/${MODEL_ID}:${GENERATE_CONTENT_API}?key=${GEMINI_API_KEY}" -d '@request.json'

The issue names no repository files or tests; start by locating the Rust Gemini integration and its request/response tests. Compare the supplied thinkingConfig request with the current integration, then determine the expected support and verify it with the relevant tests.

Written by the indexing model from the issue text.