Google touts Gemini’s 1B monthly users as it halves Gemini

Google’s August 2026 AI roundup put two trader-relevant datapoints side by side: a claim that the Gemini app surpassed 1 billion monthly users and an introductory price for Gemini 3.7 Flash set at half of Gemini 3.6 Flash’s original per‑million‑token cost. The post also positioned Gemini as a full-stack distribution push, spanning cloud models, Pixel 11 hardware running Gemini Nano, and new generative video and open-model ecosystem updates.

Key Takeaways

Gemini 3.7 Flash shipped three weeks after Gemini 3.6 Flash, with Google positioning it as a “workhorse” model for coding and agent-style workflows.
Google set Gemini 3.7 Flash’s introductory price at “half the original 3.6 Flash cost per million tokens,” but did not disclose absolute per‑million‑token pricing in the recap.
The company said the Gemini app “officially surpassed 1 billion monthly users,” and published usage stats including 63% of users talking directly to Gemini and 150 million+ images generated per day.
Pixel 11 devices unveiled at “Made by Google 2026” use Google Tensor G6 and run the latest Gemini Nano model, extending Gemini’s footprint to on-device inference.

Gemini 3.7 Flash: Faster Release Cadence, Lower Intro Token Pricing

Google’s August recap framed Gemini 3.7 Flash as a fast-follow release in its “Flash” line, landing just three weeks after Gemini 3.6 Flash. In Google’s wording, “We released Gemini 3.7 Flash as our most intelligent workhorse model yet for coding and agents.” The cadence matters because it compresses the window in which any single “efficient” model sits alone as the default choice.

The other concrete datapoint was pricing pressure. Google said 3.7 Flash launched “with an introductory price of half the original 3.6 Flash cost per million tokens,” alongside claims of “substantial improvements across software engineering, knowledge work, and web development workflows.” The post did not include the absolute dollar price for either model, which limits how precisely traders can translate “half cost” into expected margin or demand impacts.

Per‑million‑token pricing is the unit most model providers use to meter usage. Tokens are chunks of text used to measure model input and output, and pricing per million tokens is effectively a throughput tariff on inference. When a provider cuts that tariff, the immediate question is whether usage expands enough to offset the lower unit price, or whether the cut is primarily defensive in a market where model output is increasingly treated as a commodity.

Google’s 1B ‘Monthly Users’ Claim for Gemini—and the Usage Metrics It Shared

Google also used the roundup to anchor an adoption headline: “The Gemini app officially surpassed 1 billion monthly users, making it the fastest-growing product in Google’s history.” Monthly users, often tracked as monthly active users (MAU), is a count of distinct users who used a product within a month. The recap did not define how Google measured “monthly users,” what activity threshold qualifies, or what timeframe the measurement covers beyond the monthly framing.

Alongside the 1B figure, Google published a small set of usage metrics that point to how people are interacting with the product. It said “63% percent of users now talk directly to Gemini,” and tied that to a segment claim: “busy parents” are “43% more likely to use it for everyday tasks.” Google also said “Gemini now generates 150 million+ images every day,” and described small businesses as “power users” relying on Gemini’s “all-in-one image, video, and audio creation to craft marketing materials.”

For traders, the mechanism is straightforward even if the measurement is not independently verifiable from the packet. A large MAU claim can feed AI-infrastructure narratives because it implies a broad base of inference demand, while the image-generation number is a reminder that multimodal usage can be compute-heavy relative to text-only chat. The catch is that without methodology, MAU can be a wide bucket, and the market tends to overfit to the headline number.

Distribution and Product Surface Area: Pixel 11 + Gemini Nano, Omni 1.1 Flash Video, and Gemma Downloads

Google’s recap read like a distribution map, not just a model changelog. On hardware, it said: “At Made by Google 2026, we unveiled Pixel 11, Pixel 11 Pro, Pixel 11 Pro XL, and Pixel 11 Pro Fold.” The company said the lineup features “our fastest, most powerful chip, Google Tensor G6, that runs the latest Gemini Nano model.” Gemini Nano is Google’s smaller on-device Gemini model intended to run locally on phones rather than in the cloud, which can shift some workloads away from paid API calls and toward device-bound experiences.

On generative media, Google said it introduced “Gemini Omni 1.1 Flash” to provide “more precision and control for generating videos.” It listed capabilities including “scene extension, first-and-last-frame interpolation, crisp 4K upscaling, and faster prototyping.” First-and-last-frame interpolation is a technique that generates intermediate frames between a starting and ending frame to create smooth motion. 4K upscaling increases video resolution to 4K quality, often using AI to add detail. Google said Omni 1.1 Flash is available in Google Flow, Google AI Studio, the Gemini Enterprise Agent Platform, and the Gemini app.

On the open-model ecosystem, Google said: “Over one billion downloads later, Gemma supports environments from phones and edge infrastructure to space,” and pointed to a new community repository. In the same roundup, Google also highlighted climate and weather work, saying it is partnering with the UK Government and aviation leaders to expand contrail-avoidance forecasting across the North Atlantic under “Operation Blue Skies,” and that a Nature paper showed WeatherNext 2 predicts cyclone track, intensity, and wind structure with state-of-the-art accuracy, with Google open-sourcing WeatherNext 2.

Signals Traders Can Track Next: Pricing Transparency, Verification, and Feature Availability Constraints

The first signal that would turn Google’s “half cost” claim into a tradable input is absolute pricing. The recap references “half the original 3.6 Flash cost per million tokens,” but does not publish the dollar-per‑million‑token figure for 3.7 Flash or the baseline it is halving. Without that, the market is left with a directional read on pricing pressure rather than a quantifiable unit economics shift.

The second is verification and definition around the “1 billion monthly users” claim. The number is large enough to move sentiment across AI-adjacent narratives, but the packet provides no MAU methodology, activity threshold, or third-party corroboration. Even small definitional choices, like whether passive exposure counts as “use,” can change what 1B implies for compute demand.

The third is feature availability constraints on the Pixel 11 “Gemini Intelligence” surface area. Google’s recap includes qualifiers: availability is for select countries and languages, users must be 18+, features vary, and some features may require a subscription for higher usage. If those constraints loosen over time, the on-device distribution story becomes more than a marketing layer.

A final thread is whether Google expands the scope of WeatherNext 2 open-source releases or extends Operation Blue Skies beyond the North Atlantic. Those are not token-market catalysts on their own, but they are signals about how aggressively Google is willing to externalize models and tooling.

My Take: Why Google’s Pricing-and-Adoption Combo Matters More Than the Roundup Format

The part that decides this isn’t the roundup post, it’s the combination of iteration speed and unit pricing. Shipping Gemini 3.7 Flash three weeks after 3.6 Flash is a statement that the “efficient workhorse” tier is going to be competed on cadence as much as on benchmarks, and that tends to compress pricing over time.

The real test is whether Google fills in the missing primitives: absolute per‑million‑token pricing and a defensible definition for “1 billion monthly users.” If those numbers hold up and the Pixel 11 Gemini Nano surface expands beyond the current country, language, and subscription constraints, the setup starts to look structural rather than narrative-driven: cheaper tokens plus wider distribution translating into sustained inference volume.

Sources

Google Blog