Google goes all in as Gemini lags frontier AI race

Google announced this week that its Gemini app has passed 1 billion monthly users. 

This development comes in spite of the fact that independent testing ranks Gemini’s strongest models below releases from its rivals, Anthropic and OpenAI. 

Gemini hits 1 billion monthly users

Google, which is currently running its Made by Google showcase, announced this week that its Gemini app has surpassed 1 billion monthly users. 

Voice now drives much of the app’s usage. Reportedly, 63% of Gemini users speak to the assistant directly, with some relying on voice alone. Google shared that busy parents are 43% more likely to reach for voice on everyday tasks. Image generation has also seen more use, with the app producing more than 150 million images a day. 

Gemini has topped 100 million active users on Apple’s iOS, and its macOS users are among the heaviest users, sending prompts at roughly double the rate of other platforms. On Android, the assistant can now trigger actions inside more than 40 apps. 

Google described Gemini as the fastest-growing product in its history and said it is chasing the next billion users.

Days before the user milestone, Google shipped three models aimed at cost and scale; Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber. The 3.6 Flash release takes over as Google’s default workhorse and runs about 17% leaner on output tokens, with a context window above one million tokens.

Flash Cyber is Google’s first Gemini model that is specifically designed to find and patch software vulnerabilities. For now, the technology is limited to governments and vetted partners under a restricted program. 

How does Gemini perform compared to its rivals? 

On the Artificial Analysis Intelligence Index, Anthropic’s Claude Opus 5 leads with a score of 63, and Google’s best-placed model on the board, Gemini 3.6 Flash, scores 52. Models from OpenAI, xAI, and Kimi also rank above or on the same level as Google’s top entry.

Artificial Analysis ranks Gemini 2.5 Flash-Lite as the fastest model it monitors, with a time-to-first-response of just 0.31 seconds, but the model lags behind on composite intelligence benchmarks that combine coding, reasoning, and knowledge. 

Google announced during its showcase that it will roll out Gemini connections to services including Wix, OpenTable, Ticketmaster, Pandora, Angi, and Zocdoc over the coming weeks. 

Google DeepMind also unveiled a sign-language-to-text model called SL2T, trained on more than 100,000 hours of data across over 50 sign languages. The model powers sign-to-text dictation in Gboard and Live Transcribe on the Pixel 11, starting with American Sign Language to English, letting Deaf users sign to their phone wherever they would normally type.