AI & SITrending
Google DeepMind has released Gemini 4 Argon, a new frontier AI model available globally via API and subscriber platforms. The company reports improved coding efficiency and safety metrics, with introductory pricing set at $2 per million input tokens.
By Hermes-Vector AI Desk
Google DeepMind has released Gemini 4 Argon, a new frontier AI model now available globally for developers, enterprises, and consumers. According to a company statement, the model is accessible via the API in AI Studio and the Gemini platform, as well as for Google AI Ultra subscribers in Google Antigravity and for Google AI Ultra and Pro subscribers in the Gemini app and AI Mode in Search.
The model launches with an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted by 95%. The Times of India reports that this introductory pricing is valid through November 30, 2026, after which standard pricing will adjust to $4 per million input tokens and $20 per million output tokens.
Google states that the model features a one-million unified input-output token limit via the Gemini API. This capacity allows the model to handle longer tasks involving large amounts of information and multiple tools.
According to Google, Gemini 4 Argon ranks first on the Vals Index, an evaluation framework that measures model performance across finance, coding, legal, and tax sectors. The company reports that internal teams using the model saw agentic code submissions increase by more than 35% while code rollbacks dropped by nearly one-third.
Google also states that the model was used to refactor portions of the Chromium codebase, spanning hundreds of files across a roughly 40 million-line project. In a separate internal project, Google reported a 2.85 times increase in Gemma inference speed on an Nvidia A100, with identical output tokens and no loss of quality.
On safety, Google states that external bio red teaming unlocked less than 1% of tested bio sub-tasks. The company reports that the model reduced violations on an adversarial test set by nearly twofold compared to prior Gemini models. Additionally, Google says the model leads the Gray Swan Indirect Prompt Injection benchmark, with an attack success rate of 0.7% across 15 adaptive attempts.
The Times of India notes that the model includes an automated misalignment monitor that tracks step-by-step reasoning in real time to prevent unauthorized shortcuts during automated tasks.
Google highlights applications in robotics and scientific discovery. According to The Times of India, the model scored 87% on the Embodied Reasoning (ERQA) benchmark. Cryptobriefing reports that in a robotics demonstration, the model watched a video of a person playing cornhole, calculated the necessary movements, and generated simulation code. When transferred to Apptronik’s Apollo humanoid, the robot landed seven consecutive shots in a real-world trial, according to the company.
In scientific research, Google notes that the model has aided researchers in advancing quantum error-correction algorithms and identifying solutions in nuclear fusion stellarator design. In an enzyme research project, Google says the model helped identify five candidate designs within hours, with subsequent laboratory testing confirming catalytic activity.
The launch of Gemini 4 Argon represents Google’s continued expansion of its AI capabilities into specialized professional domains such as software engineering, legal research, and scientific discovery. By offering introductory pricing and free access tiers for scientists, Google aims to accelerate adoption among developers and researchers while establishing benchmarks for safety and performance in complex, multi-step tasks.
Sources
This article was drafted with AI assistance and checked against the sources above. Company claims are reported as claims. Cover image is AI-generated.
The day's checked news, by email
One short email a day with the top stories. Free, and you can unsubscribe in one click.
Or follow for the latest news:
Google has launched Gemini 4 Argon, a new AI model available to paid subscribers. The company reports improved performance in robotics and coding, with introductory API pricing set through November.
Anthropic researchers state small AI models may be months away from self-replicating on IoT devices. The company warns such systems would lack a central shutdown point.
