Gemini Hit a Billion Users. Then Google Shipped a Different One.

On 11 August 2026, Google said the Gemini app had crossed 1 billion monthly active users, according to a company blog post calling it the fastest-growing product in its history. Two days later, on 13 August, it shipped a different Gemini aimed at almost none of those billion people: a cheaper model built for coding and software agents, priced to undercut everyone else chasing the same developers.

That second launch, Gemini 3.7 Flash, is the clearest evidence yet that the company no longer treats "Gemini" as one product. It is now an umbrella over a billion-user chat app, a developer API racing rivals on price, and a scatter of smaller experiments that do not even carry the Gemini name up front. For anyone picking an AI backend, or recommending a tool to a team, knowing which Gemini you are actually getting matters more than the brand on the box.

What Actually Changed on August 13

Gemini 3.7 Flash launched at an introductory price of $0.75 per million input tokens and $3.75 per million output tokens, according to the announcement: half of the prior rate, and good only through 31 December 2026 before it rises to $1.50 and $7.50.

The release notes cite a jump from 34.4% to 43.6% on a coding benchmark and from 17.0% to 30.4% on one measuring multi-step workflow automation. The model shipped just three weeks after the tier it now undercuts on price, a turnaround attributed to developer feedback rather than a scheduled release cycle. It arrived alongside a new agent-first coding environment and developer-tool integrations, per the same post.

None of that reaches the app behind the billion-user figure. This tier is API-only, built for people writing code and agents, not people typing questions into a chat box.

The Gemini Nobody Has Heard Of

The chat app and the coding API are not the whole picture. A separate internal team has spent 2026 shipping a handful of small, distinctly named products on the same model family. Opal, a no-code builder that turns a written prompt into a working mini-app, is now folded directly into the Gemini app itself as an experimental feature, according to the company's own blog. Disco is a different animal: a standalone, invitation-only web browser for macOS that turns a user's open tabs into interactive tools, still gated behind a waitlist.

Gemini for Science pairs a hypothesis-testing agent with an internal experiment-runner, built around two papers presented earlier this year. None of the three puts "Gemini" first in its own name or marketing. Meanwhile Google Lens, arguably the highest-volume Gemini surface of all, handles image search, live translation across more than 100 languages, and shopping lookups built into everyday camera apps, with barely a mention of Gemini on its own page.

Who This Actually Helps

The 3.7 Flash pricing is aimed squarely at teams running coding agents and automated workflows at volume, where token cost compounds fast. A team running an agent that reviews pull requests all day, or one triaging support tickets on autopilot, now pays roughly half what it did three weeks earlier for a model that scores meaningfully higher on both benchmarks cited above. That shows up on the next invoice, not just in a press release.

Not every team needs that specific model, at that specific price, for that specific job. If you are still working out whether your stack calls for a cheap coding-focused tier, a general chat model, or something else, that is exactly the comparison Smart Match is built to run instead of guessing from a spec sheet.

Who It Hurts

The obvious target is DeepSeek, which built its reputation on undercutting everyone else on price for comparable coding performance. A well-funded rival now has a first-party answer at a similar rate, backed by a search business that does not need the model itself to turn a profit the way a smaller lab does. A price war against a challenger like that is one a giant can afford to lose money on for longer.

The less obvious loser is the ordinary user Google just celebrated. None of the billion people using the chat app get early access to the new coding gains, and few will ever open the smaller experiments described above. The brand keeps growing while the products under it keep splitting apart, which makes "is Gemini good at X" a harder question to answer honestly with every new release.

The Turn

The obvious rebuttal: vendors cut API prices constantly, and a 50% introductory discount on one model tier is not proof of anything beyond ordinary competition. That is fair for the price cut taken alone. It is a weaker defense of the pattern around it.

Four separately branded, independently launched products, released within the same year, none folded into a single surface, even as the company's own stated goal for its main app is to make it maximally personal and proactive for the person using it. Ordinary price competition does not usually arrive with a handful of spinout apps that almost nobody outside one internal team has used.

What to Watch

A major rival holds its own developer conference on 29 September 2026: the next moment either company could choose to consolidate its story instead of adding to it. If the smaller experiments get folded into one branded surface before then, the fragmentation was a phase. If a new standalone one ships instead, the fragmentation is the strategy.

Until then, the honest answer to "what is Gemini for" depends entirely on which Gemini someone hands you.

Frequently asked questions

What is Gemini actually used for in 2026?

It depends which Gemini. The consumer chat app, which passed 1 billion monthly users on 11 August 2026, handles conversation, image generation and voice. A separate API tier, Gemini 3.7 Flash, is priced and benchmarked specifically for coding and software agents. Several Google Labs tools, including Opal and Disco, run on the same models for narrower jobs.

Is Gemini 3.7 Flash the same product as the Gemini app?

No. Gemini 3.7 Flash is an API-only model launched 13 August 2026, aimed at developers building coding tools and agents rather than at the free consumer chat app. Its introductory rate of $0.75 per million input tokens is not something an everyday app user ever sees or pays directly.

What is Google Opal?

Opal is a Google Labs no-code tool that turns a written prompt into a working mini-app without any coding. As of August 2026 it is also embedded inside the Gemini app itself as an experimental feature Google calls a Gem, so it can be reached from either place.

Why does Google launch separate apps like Disco and Gemini for Science instead of adding features to Gemini?

Google has not stated a reason publicly. What is observable is the pattern: Google Labs has shipped several separately branded products on Gemini models in 2026, including Disco, an invitation-only macOS browser, and Gemini for Science, a hypothesis-testing research agent, none rebranded under the main Gemini name.

Is Gemini 3.7 Flash cheaper than DeepSeek's models?

Gemini 3.7 Flash's introductory rate of $0.75 per million input tokens and $3.75 per million output tokens, available through 31 December 2026, puts it in the same price band DeepSeek has used to compete on cost. Rate cards shift often on both sides, so check each vendor's current pricing before committing a production workload.

Covered in this guide

• Gemini: Google's multimodal AI model family for reasoning, coding, and creative tasks
• DeepSeek: Advanced open-source and API-based large language models with cost-effective pricing and superior reasoning capabilities.
• Gemini for Science: Gemini for Science pairs a hypothesis-tournament agent, an AlphaEvolve experiment runner, and NotebookLM review, backed by two Nature papers at I/O 2026.
• Google DeepMind: Google DeepMind, Alphabet's AI research division formed in 2023, builds Gemini, AlphaFold, and Gemma with 8,000+ researchers across six continents.
• Disco: Google Labs' experimental Chromium browser where Gemini 3 (1487 ELO on WebDev Arena) turns open tabs into no-code GenTabs mini apps in a closed macOS-only beta.
• Google Lens: Google's AI-powered visual search tool that identifies objects, translates text in 100+ languages, enables shopping via 45B product listings, and provides homework help. Free across Android, iOS, and Chrome.
• Opal: Google Labs' free no-code app builder that turns plain-English prompts into shareable AI mini-apps built on Gemini, expanding worldwide in 2026.

Sources