Silent breaking change detected in Gemini 3.5 Flash GA (May 19, 2026)

Migrated to Gemini 3.5 Flash? Your model may be dumber than yesterday.

A copy-paste model-ID swap from gemini-3-flash-preview to gemini-3.5-flash silently drops your thinking_level from high → medium. No error. No warning. Your eval scores drop quietly.

What changed silently

Before (gemini-3-flash-preview)
# Default thinking_level
thinking_level = "high"
# ~8K thinking tokens/call
After (gemini-3.5-flash) — if you don't set it
# Default thinking_level
thinking_level = "medium" ← silently changed
# ~4K thinking tokens/call

Everything you need to migrate safely

Why this matters

962 pts

Hacker News score on launch day (May 19, 2026) — 658 comments, mostly about the silent behavior change

~50%

Fewer thinking tokens per call if you migrate without setting thinking_level (high→medium default)

0 errors

The API returns 200 OK. No warning, no log line. You only find out when eval scores drop weeks later.

Fix it in 30 seconds

Add one line to your Gemini config to preserve previous reasoning quality:

# Python SDK
generation_config = GenerationConfig(
thinking_config=ThinkingConfig(
thinking_level="high" # ← add this
)
)
Full migration checklist with Python + TypeScript + REST →