Forget the leaderboard for a moment. The most consequential number in Wednesday night’s Gemini 4 Argon announcement was not a benchmark score. It was $2.
Argon launches at an introductory rate of $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off the input rate. At that launch rate, Argon costs a fifth of what OpenAI charges for GPT-6 Astra, which is priced at $10 input and $50 output. That is also half of what Anthropic charges for Claude Opus 5.5. Alphabet did not just ship a competitive model. It tried to change the entire frontier AI tier in a single evening.
The benchmarks are real enough to make rivals pay attention. But some of the specific comparisons being circulated are harder to verify independently right now. Vals AI’s public leaderboard, updated September 30, 2026, lists Gemini 4 Argon as the top model on the Vals Index, which it describes as a measure of economic impact across finance, coding, legal, and tax tasks. That said, Vals’ public page does not display the exact percentage scores cited here, and Google’s launch post does not publish those Vals Index percentages either.
But this is also where a disciplined investor pumps the brakes. Independent leaderboards do show GPT-6 Astra leading some science-oriented benchmarks and Claude Opus 5.5 leading Terminal-Bench 4.0. Self-published benchmark tables from an interested party deserve skepticism, full stop.
There is also an access problem. Argon is currently rolling out to trusted cyber defenders through Google’s Fairwind Program. Developers building products cannot yet use it broadly. The introductory price of $2 per million input tokens undercuts Astra by 5x, but it doubles when the introductory period ends. After that introductory period expires, Google says the price will be $4 per 1M input tokens and $20 per 1M output tokens, which matches Anthropic’s published pricing for Claude Opus 5.5. So the introductory price undercuts both rivals, and the regular price matches one of them. The real commodity play, if Google intends one, waits to be seen at scale.
For long-term investors, the more interesting question is structural. OpenAI is a private company burning capital without Alphabet’s search cash flows or cloud infrastructure. Anthropic relies on Amazon and Google’s own investment dollars. Neither has the distribution leverage that comes with billions of existing Google Workspace seats, Android devices, and Cloud contracts. Gemini 4 Argon is being positioned as a model for complex enterprise workflows across software engineering, knowledge work like legal and finance, and cybersecurity defense, which aligns with Google’s push to deepen its Cloud and Workspace proposition. When Alphabet prices a frontier model aggressively, it can absorb the margin compression in ways neither rival can match.
The moat question in frontier AI has never really been about which model scores highest on a given Tuesday. It has been about who controls the distribution channel when the models start to converge. Wednesday’s announcement suggests Alphabet understands that, even if the market’s after-hours reaction suggests investors are not yet convinced Argon tips the balance decisively. They may be right to wait for the general release. But the pricing signal is worth filing away.
