Gemini 4 Argon in Google Antigravity

Since the beginning, we built Google Antigravity for an agent-first future, designing foundational primitives that would scale alongside model intelligence. With Gemini 4 Argon, the agent-first experience truly comes to life in the Antigravity app.

Earlier this month, we introduced Gemini 4 Argon, our latest model with frontier performance in complex workflows across real-world software engineering, enterprise knowledge work like legal and finance, and cybersecurity defense. Today, we’re rolling out Gemini 4 Argon directly inside Google Antigravity for Google AI Ultra subscribers — available in the Antigravity desktop app and the Antigravity CLI. For businesses, Gemini 4 Argon is available via the API in Gemini platform and within Antigravity on eligible Gemini Enterprise subscriptions.

While Argon is designed to tackle your hardest problems, we’re equally excited about future models that can be used as everyday drivers for an even broader range of builders.

Accelerating development across Google

Our most honest proving ground is our engineering teams. Before bringing Argon to developers in Antigravity, we put it to work inside Google.

Today, over 100K users across Alphabet rely on Argon inside our internal version of Antigravity each month to solve complex production problems across our codebases. Over a four-week period, we have seen a 35% increase in agentic code, improving not just developer velocity, but code quality with a 30% decrease in post-submit rollbacks.

Here are a few ways teams across Alphabet have been putting Argon in Antigravity to work on our hardest engineering challenges in products you love:

• YouTube comment optimization: Argon in Antigravity helped the team cut the CPU footprint of YouTube’s comments frontend in half. By tracing production profile data, Argon identified that redundant emoji rendering alone accounted for nearly 45% of server overhead, leading to 12 targeted optimizations. The updates reduced server comment-load latency by 30 ms and dropped client end-to-end latency by 20 ms.
• Debugging Wing drone physics simulator: Argon in Antigravity reviewed Wing’s proprietary internal drone physics simulator, reasoning through the underlying physics and math rather than just the structure of the code. It highlighted areas for improved physics simulations and proposed 18 targeted modelling improvements across aerodynamics, propulsion, and structural dynamics to resolve long-standing legacy issues.

Operating at Google-scale forced us to build Antigravity to be enterprise-ready from Day 1, transforming our day-to-day development across Google. By streamlining routine tasks that every enterprise development team faces, from refactoring legacy codebases to automating test coverage and resolving build breakages, Argon is freeing engineers to focus on higher-level architecture and product innovation:

• Burning down the backlog: Argon in Antigravity took an aging queue of 134 open bug reports spanning multiple services, and ran agents to work out which were still real, cross-referencing commit histories, checking whether candidate fixes actually reached production, and inspecting live telemetry for whether the error still occurs. In under 3 hours it closed 97 stale or duplicate bugs with documented commit proof (a 72.4% queue reduction) and sorted the remaining 37 by owner with no unassigned tickets left.
• Migrating legacy code: For decades, vital scientific research has been locked inside fragile legacy codebases like Fortran. Using a custom skill within Antigravity, a team deployed Argon to map and execute a complete rewrite into modern Rust. Argon autonomously delivered the modernized codebase, achieving a 1.8x performance boost and memory safety while giving scientists a robust foundation for future work.
• Performing security audits: Argon in Antigravity was used to audit a batch of access-control changes. Argon traced how permissions actually flowed at runtime rather than just reading the code diff, and found four critical security bugs. Two earlier frontier models had audited the identical code and both concluded there were zero security holes.
• Triaging production support tickets: Argon in Antigravity batch-processes dozens of tickets from an on-call rotation. It sorts them by issue type, tracing root causes through the code and logs, proposing fixes and drafting the replies for human approval. This has freed up 2-3 hours per shift, enabling the team to focus on resolving issues faster.

We are excited to provide our Gemini for business customers the same day-to-day acceleration that we’ve been seeing with Antigravity.

Argon across disciplines

One of the most encouraging things we’ve observed internally is how quickly teams outside core software engineering have adapted Argon in Antigravity for their own use cases. Monthly active usage outside of core engineering is soaring, with over half of Cloud sales, almost half of finance, and over a third of business and operations Googlers using Antigravity. Plugins, skills, and MCP integrations make it straightforward to connect the agent to specialized data sources and workflows. Using these customizations with Argon makes it easier to integrate with deep verticalized work, such as:

• Marketing website localization: The Antigravity DevRel team used Argon in Antigravity to localize hundreds of pages of developer documentation into 17 languages in a matter of hours. This improved the accessibility and UX of the website by shrinking a multi-week process down to hours and shifting human effort exclusively to high-quality review.
• Resolving a data dashboard outage: A TPM (Technical Program Manager) used Argon in Antigravity to autonomously resolve a critical P0 data center dashboard outage that had stalled for over four months. Without needing dedicated engineering support, the TPM diagnosed deep cross-stack database failures and landed a 3,200-line pipeline overhaul, restoring real-time visibility in just a few days.
• Vetting sales prospects: Argon in Antigravity built an automated qualification workflow for an account executive in Ads sales that checks prospective company website traffic, confirms the site has clear conversion funnels, and cross-references the sales database to see whether the account is a qualified prospect. This reduced research time by 90%, saving the account executive upwards of 10 hours of manual work each week so they can spend time on direct client outreach instead.
• Financial planning and analysis: With custom finance plugins, a finance lead used Argon in Antigravity for end-to-end financial reconciliations, budget tracking, contract extraction, and daily executive briefings across 80 separate workflows. Over seven business days, 15.5 hours of human direction produced about 335 human-equivalent hours of validated financial output, saving more than eight weeks of manual work.

Getting started

Argon is rolling out starting today in the Google Antigravity desktop app and CLI for Google AI Ultra subscribers, eligible Gemini Enterprise subscriptions, and in the API in Gemini platform. Update to the latest version of Antigravity to try it on your most complex workflows. We’d love to hear what you build and where we can make the experience better.