Elon Musk's One-Word Response to Sundar Pichai's Post On New AI Model

Google's latest AI model focuses on cybersecurity, with the company claiming major gains in finding software vulnerabilities and generating secure code.

29 Oct 2026, 10:19 · NDTV

Show
Quick Read
Summary is AI-generated, newsroom-reviewed

Elon Musk praised Google's Gemini 4 Carbon AI model as "Impressive" on X
Gemini 4 Carbon targets complex software engineering and cybersecurity tasks
Google reports high scores in vulnerability detection and patching benchmarks

Why is Elon Musk praising rival AI models?

Google's latest AI model has received praise from an expected quarter.

Elon Musk - whose AI company SpaceXSI (formerly SpaceXAI) competes with the company directly - reacted with a single word of praise to Google CEO Sundar Pichai's announcement of Gemini 4 Carbon on X.

"Impressive," Musk wrote. Pichai responded to Musk's comment with a folded-hands emoji.

The brief exchange came as Google unveiled its latest frontier AI model, positioning it as a powerful tool for cybersecurity defence and complex software engineering.

Interestingly, this is not the first time Musk has responded to a Pichai post with the single word reaction - "Impressive." Back in July when Pichai posted about Google's impressive quarterly results, Musk responded with the exact same praise in one word.

What Is Gemini 4 Carbon?

Google describes Gemini 4 Carbon as its strongest model yet, with capabilities designed to tackle demanding, long-running engineering tasks and cybersecurity challenges.

According to Google's official blog, the model is built to identify software vulnerabilities, test whether they can be exploited and help develop patches to fix them. These capabilities could potentially help security teams detect weaknesses in software and respond to threats faster.

Google said Carbon achieved an 88.2% score on SEC-bench Pro, a benchmark for finding vulnerabilities, and 54.8% on ExploitGym, which evaluates exploitation capabilities. On CWE-bench, which tests vulnerability patching, the model scored 71%.

The company also said Carbon generated secure code at more than twice the success rate of its predecessor, Gemini 4 Argon, on internal benchmarks.
To be sure, these are Google's reported results, and benchmark performance does not necessarily translate directly into real-world outcomes.

Limited Rollout As Google Tests Safeguards

Google is initially making Gemini 4 Carbon available only to a select group of cybersecurity defenders through its 'Fairwind Program'.

It will gradually be rolled out to trusted partners while continuing to test and refine safeguards before a wider release, the company said.

The cautious rollout reflects an industry-wide challenge in developing AI models with advanced cybersecurity capabilities. The same tools that can help defenders identify and fix weaknesses could easily also be misused by bad actors. Rival OpenAI had stopped the release of one of its models recently over security concerns.

Google has priced Carbon at an introductory $2 per million input tokens and $10 per million output tokens, which matches the launch pricing of Gemini 4 Argon. These are charges for processing information through the model, rather than a flat subscription fee. The launch comes hot on the heels of Google introducing Argon just two weeks back.

Got a follow‑up question on this article? Go on

How may i help you today

Show full article

Track Latest News Live on NDTV.com and get news updates from India and around the world