Top AI Product

Three weeks. That’s the gap between Gemini 3.6 Flash and Google Gemini 3.7 Flash, a hosted LLM aimed squarely at coding agents and document-heavy knowledge work. No new pretraining run — Google says the gains come from algorithmic work on the core reasoning layer.

The numbers back it up. FrontierCode 1.1 Main jumps from 34.4% to 43.6%. WebDev Arena Elo goes 1538 → 1588, meaning fewer prompts to get a working layout out of a screenshot. Gemini Spark switched over to it on day one. HN put it on the front page at 480 points.

What you get through the API

Available now on Gemini API and Vertex AI: 1M token context, up to 64K output, text/image/audio/video in, and a configurable thinking level so you can dial reasoning down for cheap bulk calls and up for hard agent loops. Intro pricing is $0.75/1M input and $3.75/1M output — exactly half of 3.6 Flash — until January 1, 2027, when it doubles back to $1.5/$7.5.

The awkward part

Google has now shipped two Flash generations while Gemini 3.5 Pro stays delayed. The workhorse keeps getting faster and cheaper; the flagship still isn’t out.

You Might Also Like

• Google Gemini Spark the Gemini app Just Became a 24 7 Agent That Runs Without Your Phone
• Gemini 3 1 Flash Lite Hits ga 0 25 m Input Tokens 2 5x Faster Ttft
• Gemini Spark Leaks 5 Days Before i o Googles Agent Might buy Things Without Asking you First
• Google Nano Banana 2 Lite Gemini 3 1 Flash Lite Image a Picture in 4 Seconds for 0 034 per 1000
• Google Gemini Omni Flash Drops Video Generation to 0 10 a Second