Gemini 4 Argon: Built to solve your hardest problems

Google has launched Gemini 4 Argon, its latest frontier model, making it globally available to developers, enterprises, scientists, and consumers through the Gemini API, AI Studio, Google Antigravity, the Gemini app, and AI Mode in Search. Argon is designed for deep, multi-step work across coding, finance, legal, cybersecurity, and science, and ranks #1 on the Vals Index for performance across critical economic sectors. It launches at $2 per million input tokens and $10 per million output tokens.

Internally, Argon has driven significant engineering results: it helped the Chrome team refactor a roughly 6-million-line legacy folder within Chromium's 40-million-line codebase, sped up Google's ADK Python framework by cutting in-memory session prep 40% and cold imports 21%, accelerated the xTB quantum chemistry package by up to 2.3x, made Gemma models nearly three times faster, and enabled complex 3D visualizations. In professional workflows, it offers frontier reasoning, deep multimodal document comprehension, longitudinal data analysis, and stronger writing with less hedging and fewer clichés. External partners like Wiz used Argon to secure government data and healthcare infrastructure, going beyond detection to proactive remediation.

In science, Argon accelerated quantum error correction by translating AlphaQubit into an optimized kernel, cut spacetime costs for quantum subroutines by over 100x versus a CORDIC baseline, helped discover a broader class of counterexamples to Grad's conjecture in fusion research, assisted Gamow Labs in diagnosing a rare chromosomal rearrangement missed for years, and contributed to designing novel serine esterases. In robotics, Argon scored 87% on the Embodied Reasoning (ERQA) benchmark and enabled a "Video-to-Sim-to-Real" pipeline that let Apptronik's Apollo humanoid sink seven consecutive cornhole shots after watching a raw video. Safety features include multi-layer jailbreak defenses, agent alignment training that reduced shortcut attempts to 2%, a real-time misalignment monitor, and leading prompt injection resistance with 0.7% attack success on the Gray Swan IPI benchmark. Google is also offering 10,000 scientists free access to the highest usage tier in Google Antigravity for six months.

(Source：Gemini)