Google made its new flagship model, Gemini 4 Argon, available to developers on Tuesday through the paid tier of the Gemini API, Google AI Studio, and several other developer and consumer platforms, according to a report from The New Stack published October 14, 2026. The company first announced the model on September 30, but initially limited external access to selected cyber defenders in its Fairwind Program and other trusted testers. The model now carries competitive pricing and a unified token budget that covers both input and output.
Developers will pay an introductory rate of $2 per million input tokens and $10 per million output tokens through November 30, matching OpenAI's GPT-6.1 Sol pricing. After that date, costs will rise to $4 and $20 per million tokens, the same as Anthropic's Claude Opus 5.5. During the introductory window, Argon costs one-fifth of what Anthropic charges for Claude Fable 5.1 and OpenAI charges for GPT-6 Astra, both priced at $10 and $50 per million tokens. Cached input tokens receive a 95 percent discount during the introductory period. Argon uses a unified 1 million token limit that covers both input and output combined, meaning a long prompt can reduce how much the model can write back—for instance, loading 800,000 tokens of code leaves only 200,000 for the response. This differs from Claude Opus 5.5, which combines a standard 1 million-token context window with a separate 128,000-token output cap.
The report states that Argon ranks first on the Vals Index, which evaluates models on finance, coding, legal, and tax work, weighting each sector by its share of U.S. GDP. On Embodied Reasoning Question Answering (ERQA), a test measuring physical and spatial reasoning, Argon scored 87 percent, outperforming Claude Opus 5.5 at 82.3 percent, GPT-6 Astra at 81 percent, and Claude Fable 5.1 at 74 percent. Bloomberg reported September 30 that some Google employees with access said the model struggled with certain coding tasks, particularly front-end design, though Google told Bloomberg it would be "inaccurate to say that Gemini 4 is underperforming in areas such as coding." Artificial Analysis published an evaluation on September 30 showing Argon scored 53 on its Intelligence Index, matching GPT-6 Astra at tested reasoning settings, though there's no guarantee the model released Tuesday is identical to the one tested then.
Google's strategy centers on making Argon useful for long, multistep tasks in coding, science, and agent workflows by providing extra token capacity, according to the report. More than 100,000 Google employees use Argon each month in an internal build of Antigravity, the company's platform for running coding agents, and over four weeks, code submitted by agents rose 35 percent while post-submit rollbacks fell 30 percent. The company says more than half of its Cloud sales staff and almost half of its finance team now use Antigravity. Google trained Argon to check with a person before doing anything it can't undo and to report when it hits a wall instead of working around it, with the model trying unauthorized shortcuts 2 percent of the time on coding evaluations—the lowest rate for any Gemini model, though the company acknowledges that when thousands of agents work together, 2 percent still represents quite a few agents that could go rogue.
Argon is available in Antigravity for macOS, Windows, and Linux desktop apps and in the Antigravity CLI, but only for Google AI Ultra subscribers and eligible Gemini Enterprise plans, while AI Pro subscribers get Argon in the Gemini app but not in Antigravity. Google is also expanding Fairwind, the program launched September 2 to give vetted security teams early access to its cyber models, adding hundreds more partners to the more than 650 the program had at launch—selected Fairwind partners can access Argon without its cyber guardrails. On Gray Swan's indirect prompt injection benchmark, Argon leads with attacks succeeding 0.7 percent of the time across 15 adaptive attempts. Developers have until November 30 to test Argon at the introductory rate before pricing doubles to match Anthropic's premium tier. The tiered access model and aggressive introductory pricing suggest Google is betting on developer adoption momentum to establish Argon before competitors can respond, while the gap between Pro and Ultra subscription features may push serious users toward higher-cost plans once the initial discount expires.