Google Unveils Gemini 4 Argon AI Model After Delays

SAN FRANCISCO — After months of delays and a shake-up at its flagship AI lab, Google is back in the race. Alphabet’s Google on Wednesday unveiled Gemini 4 Argon, its new flagship artificial intelligence model — the largest and most capable model the company says it has ever built, designed to compete head-on with OpenAI’s Astra and Anthropic’s Opus.

The announcement, made Wednesday afternoon, marks the start of Google’s “Gemini 4” generation of models and arrives as the company scrambles to close the gap with rivals that have surged ahead during Google’s prolonged development stall. CEO Sundar Pichai shared the news on social media, saying the model delivers “frontier performance in complex workflows across real-world software engineering, knowledge work, and cybersecurity defense with an industry-leading 1M token output limit,” according to Reuters.

What is Gemini 4 Argon?

Argon is a full-size flagship model — larger than any of Google’s previous “Pro” tier models, a company spokesperson told Reuters. Google describes it as its most performant model yet for complex workloads, and the company says it compares favorably with frontier models from OpenAI and Anthropic on key coding and cybersecurity benchmarks.

The standout specification is the industry-leading 1 million-token output limit, which lets the model generate far longer and more complex responses in a single pass — a meaningful advantage for software engineering, legal work, financial analysis, and cybersecurity defense, the domains Google is targeting first.

Pricing is another weapon: Google set Argon at $2 per million input tokens and $10 per million output tokens, significantly undercutting competitors, according to Blockonomi. For developers and enterprises running models at scale, that price gap could prove decisive.

Gemini 4 Argon benchmarks: strong, but not dominant everywhere

Google’s own release materials show Argon posting better marks than Astra and Opus on several self-reported industry benchmarks. But the company also acknowledged areas where it still trails — Argon remained behind on two of the four coding-related benchmarks Google included in its release, Reuters reported.

The mixed benchmark picture reflects Google’s new messaging. Rather than claiming outright supremacy, the company has shifted toward emphasizing cost advantage and real-world performance on complex workflows — a pragmatic pivot after losing ground during the delay.

Who gets access — and when

Argon is not available to the general public yet. Google is rolling it out in stages, starting with trusted cybersecurity partners through its Fairwind Program, paying API customers, and Google AI Ultra subscribers, according to CoinCentral and Blockonomi. The company is also participating in the Trump administration’s voluntary process for pre-release access to frontier AI models, giving government reviewers early access.

Google has not provided a timeline for broader public release. Developers, enterprise customers, and general consumers will get access in subsequent phases, the company said.

The delays that cost Google its lead

Argon’s arrival closes a chapter the company would rather forget. Google had originally planned to release a Gemini 3.5 Pro model in June, Pichai said earlier, but that release was scrapped entirely, according to Reuters. The delay cascaded through Google’s AI roadmap as the company overhauled its DeepMind lab — founder and CEO Demis Hassabis stepped aside, and several Gemini team leaders left the company.

Meanwhile, rivals kept shipping. Anthropic pushed new Opus releases and OpenAI advanced its Astra line, extending the gap while Google regrouped. The months of silence fueled doubts about whether Google could still compete at the frontier — doubts that Wednesday’s announcement is meant to dispel.

Wall Street’s verdict was cautiously positive: Alphabet shares gained roughly 1% in regular trading Wednesday, added about 2% in after-hours and premarket trading Thursday, according to Blockonomi, suggesting the market welcomed the news without fully declaring Google’s comeback.

What Gemini 4 Argon means for the AI race

Argon reframes Google’s strategy. Rather than trying to win a pure capability race it is still chasing, Google is competing on a different axis: lower cost, longer outputs, and real-world utility. The aggressive pricing undercuts both OpenAI and Anthropic and could win over developers at scale.

Still, the AI race is now a marathon of shipping cadence, not a single launch. Whether Argon marks a true turning point for Google’s AI ambitions depends on what follows: how fast the company gets the model into developers’ hands, how quickly it can follow with the next iteration, and whether the new leadership at DeepMind can rebuild the momentum it lost.

For now, the message is clear: Google is back at the frontier — even if it’s arriving from behind.