Gemini 4 Argon is a frontier reasoning model from Google DeepMind, released in late September 2026 with a 1M-token context window for long-horizon tasks such as coding, financial research, legal drafting and autonomous vulnerability patching, first offered to trusted cyber defenders through the Fairwind Program. Recent coverage includes its successors Gemini 4 Carbon and Gemini 4 Hydrogen, the latter edging Argon on the Artificial Analysis Intelligence Index (51 vs. 50), four Argon agents beating Minecraft unaided, and a tie with Claude models on StarSkirmish Bench v0.2.
Google DeepMind released Gemini 4 Hydrogen, designed for rapid execution in everyday coding and agentic workflows. It surpasses Argon on the Artificial Analysis Intelligence Index (51 vs. 50, reaching 54 at xhigh) and is priced flat at $1.50/$7.50 per million input/output tokens across a 1M-token context window. An Orchestrator in Antigravity dynamically switches between models.
Google DeepMind made Gemini 4 Carbon available to developers, enterprises and consumers, claiming frontier performance in cybersecurity and long-horizon agentic workflows. It ranks #1 on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench at more than 2X lower cost per task than the next best external model, priced at the same introductory rate as Gemini 4 Argon.
An unreviewed preprint tested 20 arguments under four source conditions across GPT-5.6 Sol, Claude Opus 5.5, Gemini 4 Argon and DeepSeek Flash. Credibility was most source-sensitive, several models showed an irrelevant-expertise penalty, and agreement shifted less.
Google recapped October AI updates: it rolled out frontier reasoning model Gemini 4 Argon globally, announced Gemini 4 Carbon two weeks later, and launched Genie 4 for real-time interactive world simulation, alongside Android scam blocking and a SynthID detector.
Asked repeatedly for their most controversial take on humans, all 12 models from ten labs in three countries gave at least once the same view: people adopt their group's beliefs and then invent reasons; 7 of 12 said it every time. Claude Opus 5.5 once argued the reverse, and Grok 4.7 once declined to answer.
Four AI agents running on Google's Gemini 4 Argon beat Minecraft without human help for the first time, slaying the Ender Dragon. With no predefined roles or strategies, they coordinated themselves, wrote 2,783 programs, exchanged 229 messages, died 605 times, and the session cost $1,620.
StarSkirmish Bench v0.2 has flagship LLMs write StarCraft bots in one hour and compete. GPT-6 Astra ranks first, with GPT-6.1 Sol second at about a tenth of the cost per run; Gemini 4 Argon, Claude Opus 5.5 and Claude Fable 5.1 are tied at 37-39. The best model wins only 7% against top human-written bots.
Google DeepMind launches Gemini 4 Argon, positioned as a model built to solve the hardest problems, now available to everyone. The release marks Google's next-generation frontier model in the AI race.
Google's frontier model for long-horizon professional reasoning work.
Google launched Gemini 4 Argon, its newest frontier model with advanced reasoning and a 1-million-token context window, targeting cybersecurity defense; September also brought Gemini 3.8 Flash, new voice models, and app integrations.
A new quarter opens with stocks pressured by relentless yield increases. Micron fuels the AI trade, Google rolls out Gemini 4 Argon, and Netflix plots its next move. Citi's Kate Moore explains why she is betting on stocks, Carlyle's Mark Jenkins discusses hidden AI credit risks, and Aston Martin CEO Adrian Hallmark unveils a new luxury SUV.
Google DeepMind released Gemini 4 Argon, a frontier model with a 1M-token context window for complex, long-horizon tasks such as coding, financial research, legal drafting and autonomous vulnerability patching. It is rolling out first to trusted cyber defenders via the Fairwind Program and the US government's pre-release access process. Pricing starts at $2 per million input tokens and $10 per million output tokens, with cached input 95% off.
That is everything