July 21, 2026
Too hot to handle, too ignored to care
Gemini last models: temperature, top_p, and top_k are deprecated and ignored
Google tells Gemini users: your favorite knobs do nothing now, and commenters are split
TLDR: Google’s new Gemini models are here, but the biggest news is that old tuning controls now do nothing and may soon cause errors. Commenters are split between "good riddance" and "why replace simple settings with longer instructions," turning a routine update into a mini-drama about who really controls AI output.
Google just dropped two shiny new AI models — Gemini 3.6 Flash and Gemini 3.5 Flash-Lite — but the real fireworks started when people noticed a blunt little warning: the old creativity controls, basically the sliders developers used to tweak how wild or predictable the AI sounded, are now deprecated and ignored. Even juicier? In future versions, sending them at all could trigger a 400 error, meaning your request may simply get rejected. So yes, one of the biggest takeaways from the launch was not "look at the cool new model," but "your favorite buttons are fake now."
And the comment section immediately turned into a tiny civil war. One camp basically cheered, with one user declaring these settings had been "basically useless" for recent models and often made results worse. Another crowd was far less convinced, side-eyeing Google’s advice to use stricter written instructions instead. One commenter mocked that idea as sounding like a smaller version of an Agent.md file — in plain English, a giant note begging the AI to behave. Translation: people are not thrilled about replacing simple controls with more prompt babysitting.
Then came the bigger theory-posting. A few commenters connected the move to a wider industry trend, noting even rival models are dropping the same controls. The vibe was half detective board, half shrug emoji: maybe modern AI is getting more powerful, but also more fragile. The funniest part? This was supposed to be a product update, and the community turned it into a debate over whether the steering wheel was ever attached in the first place.
Key Points
- •Gemini 3.6 Flash and Gemini 3.5 Flash-Lite are generally available and positioned as production-ready Gemini API models.
- •Gemini 3.6 Flash is priced at $1.50 per 1 million input tokens and $7.50 per 1 million output tokens, while Gemini 3.5 Flash-Lite is priced at $0.30 and $2.50 respectively.
- •Both models support a 1 million token context window, up to 64,000 output tokens, thinking, and built-in tools including Computer Use.
- •The migration guide says developers must remove deprecated sampling parameters `temperature`, `top_p`, and `top_k`, along with prefilled model turns.
- •Gemini 3.6 Flash is now the default model for the Antigravity agent in Gemini Managed Agents, with an API field available to change that setting.