Gemini 4 Argon

Gemini 4 Argon is the frontier model Google announced on 30 September 2026 in a post by Koray Kavukcuoglu, SVP of Google DeepMind and Google’s chief AI architect. Google describes it as built for long-horizon work in software engineering, enterprise knowledge work such as legal and finance, and cybersecurity defence, and says it already runs in Google’s internal workflows. At release it is rolling out only to a set of trusted cyber defenders through Google’s Fairwind Program. Google says it is taking part in the US government’s voluntary process for pre-release model access, will widen access gradually, and will release it next to developers, enterprises and consumers, starting with paid API customers and Google AI Ultra subscribers, without giving a date. The post gives an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input 95% cheaper, and says $4 and $20 apply after the introductory period; the board records the later rate. Google’s post does not state a model identifier, context window or knowledge cutoff, and the Gemini API model, pricing and changelog pages did not list the model when read on 1 October. Artificial Analysis lists a 1 million-token context window with text and image input and text output. The post describes safeguards against misuse for cyber and CBRN attacks, monitoring of the model’s chain of thought and actions for misalignment, and leading results on Gray Swan’s indirect prompt-injection benchmark; Google’s own evaluation figures are its claims and are not stored as scores. The Verge reported the release independently the same day. Artificial Analysis, measuring independently, scores it 53 on version 4.3.2 of its Intelligence Index at its High setting and ranks it eighth of 223 models, against 52 for GPT-6.1 Sol at maximum effort and 58 for Claude Opus 5.5 in the same version.