Gemini API: Gemini 3.6 Flash GA, Gemini 3.5 Flash-Lite GA, and sampling parameter deprecations
Google’s Gemini API changelog says Gemini 3.6 Flash and Gemini 3.5 Flash-Lite are now generally available, and that temperature/top_p/top_k are deprecated.
Editorial Team
The Authority AI Tools editorial team maintains this directory using vendor documentation, dated source checks, product changelogs, and clearly identified hands-on observations where available.
Google’s Gemini API changelog says Gemini 3.6 Flash and Gemini 3.5 Flash-Lite are now generally available (GA).
New GA models
In an entry dated July 21, 2026, Google lists two newly GA Flash-family models:
- Gemini 3.6 Flash (
gemini-3.6-flash) — Google says it has improved token efficiency and better code/agentic planning capabilities, with a lower price point than 3.5 Flash. - Gemini 3.5 Flash-Lite (
gemini-3.5-flash-lite) — Google positions it as a low-latency, cost-effective option “designed for high-volume automation.”
If you run tool-using agents, the “Flash-Lite” tier is often a good default for subagents (triage, extraction, lightweight checks), while a stronger Flash model can handle planning or complex tool orchestration.
Deprecated parameters: temperature, top_p, top_k
The same July 21 entry also says the sampling parameters temperature, top_p, and top_k are now deprecated.
If your code sets these explicitly, plan time to review Google’s latest model guidance and update configurations.
Source
- Google AI for Developers — “Gemini API: Release notes” (July 21, 2026 entries): https://ai.google.dev/gemini-api/docs/changelog
Free Resource
2026 AI Coding Tools Comparison Chart
Side-by-side comparison of features, pricing, and capabilities for every major AI coding tool.
No spam, unsubscribe anytime.
Frequently Asked Questions
What is Gemini API: Gemini 3.6 Flash GA, Gemini 3.5 Flash-Lite GA, and sampling parameter deprecations?
Related Articles
Cursor open-sources Mixture-of-Kittens, an MoE training megakernel
Cursor Research says its open-source Mixture-of-Kittens megakernel for GB300 NVL72 systems fuses MoE communication and computation and delivered a 1.41× end-to-end tokens-per-second gain in its production stack.
Read more →Product UpdatesCursor adds Google Workspace plugins for agents
Cursor’s August 3, 2026 changelog adds Google Workspace plugins that let coding agents read, write, and act across Drive, Gmail, Calendar, Docs, Sheets, and Chat.
Read more →Product UpdatesClaude Code 2.1.224 adds self-hosted runners and cross-session messaging
Anthropic’s official Claude Code changelog update for August 7, 2026 adds self-hosted runners for web, mobile, and desktop sessions plus cross-session messaging and other workflow changes.
Read more →