Gemini 4 Argon: Built to solve your hardest problems

Google’s AI blog post introducing Gemini 4 Argon repeats the core launch: availability for developers and enterprises via API, Ultra access in Antigravity, and Pro/Ultra access in the Gemini app and AI Mode in Search. The company highlights deep work across coding, finance, legal, cybersecurity, and science, #1 Vals Index ranking, and introductory pricing of $2 / $10 per million input/output tokens.

Cited impacts include internal engineering acceleration, Isomorphic Labs drug-discovery workflows, Wiz cyber findings on government and healthcare systems, and scientific results from quantum to fusion counterexamples to rare disease and enzyme design. Robotics ERQA performance and a Video-to-Sim-to-Real cornhole demo with Apptronik’s Apollo are featured.

Safety sections summarize misuse defenses, misalignment monitoring, and leading Gray Swan IPI results, with pointers to the model card and full safety report.

Two near-duplicate launch posts—blog.google and DeepMind—exist because Google needs both consumer distribution and research credibility. The substance is the same bet: Argon as the model you give hard, multi-step jobs, not small talk.

We read the science and cyber anecdotes as invitations to verify, not as settled history. A chromosomal find missed for years, enzyme designs confirmed in lab, hospital bugs caught by Wiz—each is important if independently documented. Until then, they are directed testimonials.

Price and access will drive near-term behavior more than any single demo. Teams will try Argon because it is cheaper to experiment than several rivals on paper. Whether they stay depends on agent reliability, timeout rates, and how the 1M window behaves when prompts eat the budget.

Google is back on the field. The season is evaluations, not blog posts.

Google is synchronizing marketing across consumer and DeepMind channels for a major Gemini 4 family launch starting with Argon.