What happened
VentureBeat reports that Google has expanded Gemini 4 Argon from limited early testing to broader access through the Gemini API, Google AI Studio, Antigravity, eligible Gemini Enterprise subscriptions, and selected Gemini consumer plans. Google is offering introductory API pricing through November 30, after which standard prices will double.
VentureBeat reports that Google made Gemini 4 Argon broadly available on October 13, 2026, 13 days after initially limiting it to approved cybersecurity defenders in the Fairwind Program. The model is available through the paid Gemini API, Google AI Studio, Antigravity, and eligible Gemini Enterprise subscriptions. Google AI Pro and Ultra subscribers can access it in the Gemini app and AI Mode in Search, while Antigravity access for individuals requires the Ultra tier. The source also reports availability through Android Studio and Firebase AI Logic, but the supplied material does not independently verify the rollout or specify every access restriction.
Google is offering introductory API prices of $2 per million input tokens, $10 per million output tokens, and $0.10 per million cached input tokens through November 30, 2026. Beginning December 1, standard prices rise to $4 per million input tokens and $20 per million output tokens. VentureBeat reports that Google describes a combined input/output limit of up to 1 million tokens through a specialized continuation protocol, but says Google has not clarified how that limit relates to a single uninterrupted response.
Google cites internal examples involving Chromium code restructuring, Agent Development Kit optimization, Gemma , bug triage, cybersecurity and scientific research. It reports that more than 100,000 employees use an internal Argon version monthly and claims a 35% increase in agentic code submissions alongside 30% fewer rollbacks over four weeks. These are company-reported figures, not independently replicated enterprise evaluations.
The report also cites contrasting external assessments. Vals AI ranked Argon first on its GDP-weighted index as of October 12, while Artificial Analysis placed it level with some competitors and below others. VentureBeat’s independent reviewer described the model as highly capable but not the best overall, and criticized Antigravity as trailing established coding tools. No source-provided evidence establishes how Argon performs on a particular customer’s workload.
Source details: venturebeat.com ↗
Why it matters
The release gives developers and enterprises a new Google model to evaluate for coding, analysis, cybersecurity, scientific work, multimodal tasks and agentic workflows. Its practical significance will depend on reliability, tool integration, token consumption and performance on customers’ own systems—not aggregate benchmarks alone. VentureBeat reports strong company-reported results and mixed independent evaluations, while noting that the claims have not been independently replicated as enterprise outcomes.
Argon’s importance is its combination of model access, Google product integration and a limited introductory price window. Enterprises already using Google Cloud, Workspace or Gemini can evaluate the model within a broader platform rather than adopting a standalone model service. That could matter even if Argon does not lead every .
The reported capabilities span software engineering, visual understanding, agentic task execution, cybersecurity, scientific computing and robotics. If independently validated, that breadth could make Argon useful for organizations seeking one model across several workflows. However, the source does not establish general reliability, uptime, customer support terms, data-handling conditions or production readiness for high-consequence use.
Token prices alone do not determine total cost. The source notes that models can consume different amounts of reasoning tokens and require different numbers of attempts. Buyers will need to measure completion rates, latency, tool costs, review requirements and failure recovery on their own workloads.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
What is a common training objective for an autoregressive language model?
What to watch next
Watch real-world enterprise evaluations, access limits across Google products, the effect of the December pricing increase, and whether Argon’s extended reasoning configuration delivers enough value to offset potentially high token use. Google’s reported safety and productivity results also require independent scrutiny before organizations rely on the model for consequential actions.
The most immediate issue is whether the December 1 price increase changes the economics of production deployments. Organizations evaluating Argon during the introductory period should model costs at the post-November rates rather than treating the discount as a lasting price.
Independent testing should examine Google’s claims about coding productivity, vulnerability remediation, safety behavior, scientific results and robotics. The reported 0.7% attack success rate on one prompt-injection and other safety figures do not remove the need for access controls, monitoring and human approval.
Developers should clarify how the advertised 1-million-token API limit works in practice, including continuation behavior, latency, context handling and billing. The supplied report leaves those details unresolved.
Google has not announced a public successor date, although VentureBeat reports that Business Insider described internal testing of updated Gemini 4 variants. That report is not independently confirmed here and should not be treated as a product roadmap.