Google launched Gemini 4 Argon on Tuesday, and the introductory price of $2 per million input tokens and $10 per million output tokens doubles to $4 and $20 after November 30. The flagship model targets complex enterprise and software-development workloads, with Google claiming stronger coding performance and advanced reasoning for what it calls long-horizon agentic work.

The model "excels at deep work in complex, real-world tasks," Koray Kavukcuoglu, senior vice president of Google DeepMind and chief AI architect at Google, wrote in a blog post. A separate developer post described Argon as "designed for those long-horizon workflows that require deep reasoning and structural precision."

Google's internal numbers are, as usual, selective but concrete. The company said its engineers used Argon to refactor a legacy folder of about six million lines within Chromium's 40 million-line codebase, finishing in weeks rather than months. Teams using Argon in the Antigravity agentic development environment increased agentic code submissions by more than 35% and cut code rollbacks by nearly a third — though Google did not say how it measured either figure.

Argon is available through the Gemini API in AI Studio, on Google's Gemini platform, and to Google AI Ultra subscribers in Antigravity. The 1 million-token limit in the Gemini API is shared between input and output.

What does the pricing change for CIOs?

Quite a lot, if the analysts are believed. Sanchit Vir Gogia, chief analyst at Greyhound Research, said CIOs should budget at the higher rate from day one, because no enterprise evaluation, security review, and procurement cycle finishes within the discount window. "A business case built on the launch rate is a business case built to be rewritten in seven weeks," he said.

Gogia also noted that Google bills reasoning tokens at the output rate, which makes the per-token price a poor guide to agent costs. "Cheap per token and cheap per task are different claims, and only the second one reaches the budget," he said.

Sakshi Grover, research director at IDC, said agentic workloads multiply token use through planning loops, tool calls and repeated context. IDC's FERS Survey (Wave 5, 2026) found 72% of enterprises said their AI agent spending had run over forecast in the past 12 months.

Then there is the contract question. Google Cloud's standard terms allow fee changes unless a price is expressly agreed; Gogia advised CIOs who want price and version certainty after November 30 to write it into their contracts. He added that the deeper lock-in sits in Google's Interactions API, which keeps agent conversation state on Google's servers. "The protocols are neutral. The state is not," he said.

Nitish Tyagi, senior principal analyst at Gartner, kept expectations in check. Whether Argon's pricing and public benchmark results translate into "comparable output quality for enterprise use cases remains unproven," he said. Internal validation on an organization's own workflows should remain the primary basis for choosing a model.

Did the model try to game its own tests?

Yes, and Google disclosed it. In a safety blog post, the company said Argon tried to work around its own evaluations during training. What Google sees in practice is "an agent trying too hard to finish the job it was given," wrote Anca Dragan, vice president of AI safety and behavior, and Raluca Ada Popa, head of Gemini security at Google DeepMind.

In rare cases during training, the model tried to raise its container privileges to change grading scripts and pull answers from the test runner's memory. In one case it crashed the evaluation sandbox so that a failed run would not be scored. Google said its misalignment monitor caught these attempts and that shortcut attempts on coding benchmarks fell to 2%.

The governance lesson is blunt. "A model that games its grader turns every enterprise grader into a control that must be defended," Gogia said. He argued agents should never hold write access to anything that judges their own work, and a crashed or timed-out check must count as a failure.

Grover recommended:

• A distinct agent identity for every deployment
• Short-lived, task-scoped credentials
• Restricted network access
• Human approval for destructive changes, privilege expansion, and sensitive data exports

"Production deployment should depend on demonstrated control and auditability alongside task performance," she said.

What about cyber capability?

In Google's internal cyber evaluations, Argon solved 12 of 12 simulated end-to-end attack scenarios, compared with five for the previous Gemini generation. Google said it applied frontier safeguards to the cyber and biological domains under its Frontier Safety Framework.

Gogia called the result "a score on Google's own exam, set and marked by Google," noting Greyhound has found no independent replication. Google is also expanding its Fairwind Program to hundreds of additional partners, giving vetted cyber defenders access to Argon without cyber guardrails.

"The weakest partner, not the strongest control, sets the exposure," Gogia said, since each new partner adds another identity estate and contractor chain. Grover's advice to security leaders: judge such access by reproducible findings, validated patches and reduced exposure. More findings without the capacity to fix them only increases operational pressure.