Gemini 4 Argon is now available worldwide as Google's most capable frontier model

Google began rolling out Gemini 4 Argon globally today, making its new flagship model available to developers, businesses and consumers. Argon is reachable through the API in AI Studio and the Gemini platform. Google AI Ultra subscribers can use it in Google Antigravity, the company's agentic development environment, while Ultra and Pro subscribers get access through the Gemini app and the AI mode in Google Search.
The rollout is tiered in a way that changes what lower-paying users see. Free Gemini users and paying Plus subscribers now have access only to the weaker Flash models under a new model tier structure. Argon takes over the Gemini Pro class as Google's frontier series. Google says additional Gemini 4 models will follow, which points to refreshed Flash variants that would eventually reach free and Plus tiers. Google is also reportedly already testing an internal successor codenamed Carbon, intended to beat Argon specifically at coding.
Pricing holds at the previously reported introductory rate: $2 per million input tokens and $10 per million output tokens. Cached inputs carry a 95 percent discount, which Google describes as the cache being 95 percent cheaper than the regular input token price rather than stating the effective per-token figure directly.
On benchmarks, the picture is mixed. Argon holds the top spot on the Vals Index but still trails Anthropic's Claude Opus 5.5 in independent overall rankings. Because Google waited roughly two weeks to ship the model, it now also sits behind Meta's Muse Wonder Full, which launched just yesterday.
Google published internal numbers alongside the launch, and these are self-reported. Software engineers using Argon increased agent-driven code submissions by more than 35 percent, the company says, while code rollbacks dropped by nearly a third. In the Chromium project, Argon helped the Chrome team complete an architectural refactor spanning hundreds of files inside the 40-million-line codebase. Work that would have taken months finished in a few weeks. The user-facing payoff is that web pages can move between tabs, sidebars and floating windows without reloading.
Working with Google's open-source Agent Development Kit, Argon ran autonomously overnight and found optimizations that cut session preparation time by 40 percent and sped up cold imports by 21 percent. Inference speed for the open Gemma models on an A100 via vLLM rose by a factor of 2.85, from roughly 37,900 to over 108,000 tokens per second. Those changes are visible in upstream pull requests, which makes them more checkable than the aggregate productivity claims.
The model can also generate complex 3D models, games and animations as pure code that runs directly in the browser. One demo is a raytracing engine written entirely in Python with no external libraries, and Google published the source code. In another demo, Argon starts from a single image inside Antigravity and controls Blender via MCP in an autonomous loop, modeling gothic architecture and setting up audio-reactive lighting to produce a finished 3D scene. MCP here refers to the Model Context Protocol, a common way for models to drive external tools. Google notes that 3D remains a weak spot for Meta's Muse models, which Meta AI chief Alexandr Wang said were not specifically trained for 3D demo games.
Google is extending access through several programs. It is offering 10,000 researchers the highest usage tier of Google Antigravity free for six months, awarded on a rolling basis depending on eligibility, capacity and selection, with applications open now. Cybersecurity defenders can apply separately to the Fairwind program, which grants access to Argon without cyber safeguards; Google plans to expand it to hundreds of additional partners.
On safety, Google calls Argon's architecture its strongest to date. External red teaming in the bio domain bypassed fewer than 1 percent of tested safety checks, and on an adversarial set violations dropped by nearly half compared with earlier Gemini models. Attempts to take unauthorized shortcuts on agent alignment fell to 2 percent, with an automated misalignment monitor analyzing the agent's reasoning and actions in real time. Argon also posted the lowest attack rate on the Gray Swan Indirect Prompt Injection benchmark, a successful attack rate of just 0.7 percent across 15 adaptive attempts. For comparison, on the previous Gray Swan test GPT-6 Astra hit 8.5 percent and Claude Opus 5 came in at 4.8 percent. Google's write-up does not make clear whether the exact same tests were used, so treat that comparison as approximate. More safety detail is in the model card.
For developers, the practical considerations are price, placement and workflow fit. At $2 and $10 per million tokens with a deep cache discount, Argon is positioned as an aggressively priced frontier option, so applications with heavy prompt reuse should see the most benefit from caching. The tier restructuring means free and Plus users lose access to the class of model they previously had, which could affect apps built for that audience. Availability across AI Studio, the Gemini platform, Antigravity, the Gemini app and Search AI mode means teams can prototype and ship against the same model without switching providers, though benchmark leadership is contested and self-reported productivity figures deserve independent verification. The promised follow-on Gemini 4 releases and reported Carbon work signal that Argon's position at the top of Google's lineup may be short-lived.
Why it matters: Teams building on Google's stack can now target a single frontier model across API, app and agent tooling, with pricing that undercuts many rivals and a cache discount that rewards prompt-heavy workloads. Independent rankings still favor Claude Opus 5.5 and, as of yesterday, Meta's Muse Wonder Full, so model choice remains a benchmark-by-benchmark decision rather than an obvious default. The tier change is the quieter disruption: free and Plus users are now on Flash models, and any product depending on their access to frontier capability will need rethinking.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Google is making Gemini 4 Argon available worldwide through its API, the Gemini app, and Google Antigravity. The frontier model is aggressively priced but trails Anthropic's Claude Opus 5.5 and Meta's Muse Wonder Full in independent benchmarks. Internally, Google says Argon already cut coding tasks from months to weeks. The article Gemini 4 Argon is now available worldwide as Google's most capable frontier model appeared first on The Decoder .