Google ran into trouble after the rocky rollout of its latest AI agent, and is now trying to patch things up fast. The mixed reception to Gemini 3.5 Pro (which was delayed several times and, according to reports, didn’t meet internal targets) seems to have pushed the team toward a damage-control sprint.
Three weeks elapsed between Gemini 3.6 Flash and Gemini 3.7 Flash — an unusually short gap. That pace suggests Google is betting on the Flash line to quickly change perceptions, or at least to blunt the fallout from the earlier misstep.
A few concrete points (human abbreviations included): promo pricing thru end-2026 is $0.75 / 1M input tokens and $3.75 / 1M output tokens. The output fee is half what Gemini 3.6 Flash charged ($7.50 / 1M tokens). Performance also moved: Agent Arena ranking climbed to 20 (vs. 35 for the prior Flash), and DeepSWE scored 65.5% — about +18.8% over 3.6 Flash. So, better scores and a cheaper price, at least on paper.
Whether this rapid cadence and the price cut amount to a durable rebound is another question; quick fixes can look good in benchmarks and press releases, but they don’t always hold up in real-world deployments. Still, for now Google has a fresher model, lower per-token costs, and some measurable gains to talk about.