{"rewrite":{"id":"r_5ba0701c73cf49a4a8043674","clusterId":"c_8c4d62532acd5fca8249d505","slug":"google-announces-gemini-4-argon-frontier-model-with-1-million-token-output","model":"deepseek-v4-1-flash","headline":"Google Announces Gemini 4 Argon Frontier Model With 1 Million Token Output","summary":"Google announced Gemini 4 Argon on September 30, 2026, the first model in its Gemini 4 series. Google published benchmarks showing Argon leading on knowledge work, long-context processing, and business automation. On the Vals Index, Argon scored 68.9% against GPT-6 Astra's 63.1% and Claude Opus 5.5's 67.0%. On GraphWalks in the 256,000 to 1 million token range, Argon scored 84.2% against Astra's 71.8% and Opus 5.5's 66.8%. On AutomationBench, Argon scored 51.3% against Opus 5.5's 42.5% and Astra's 41.4%. Google did not release Argon to the public. It is being provided through the Fairwind Program to vetted cyber defense organizations and thousands of Google employees. Google plans to expand access in stages to paid API customers, Google AI Ultra users, developers, enterprises, and general users. Introductory pricing is $2 per 1 million input tokens and $10 per 1 million output tokens, rising to $4 and $20 after the introductory period.","whyItMatters":"Google's own benchmark table is not an independent evaluation, and Argon trails competing models on coding and OS tasks like FrontierSWE v2 and OSWorld-2.0, so the performance lead is conditional on which benchmark is being read.","webCardHtml":"\u003cp\u003eGoogle CEO Sundar Pichai said on X that \u0026#34;with discussion building around the next model, we wanted to preview it as soon as possible.\u0026#34; Several thousand Google employees already use Argon for coding, research, and writing.\u003c/p\u003e\n\u003cul\u003e\n  \u003cli\u003eDeepSWE v1.1 (long-duration software development): Argon 77.9%, GPT-6 Astra 74.1%, Claude Opus 5.5 74.2%, Claude Fable 5.1 67.4%.\u003c/li\u003e\n  \u003cli\u003eVals Finance Agent v2: Argon 65.4%, Fable 5.1 58.9%, Opus 5.5 58.6%, Astra 53.5%.\u003c/li\u003e\n  \u003cli\u003eHarvey Legal Agent Benchmark: Argon 19.6%, Astra 5.4%, Opus 5.5 3.8%, with all models still low in absolute terms.\u003c/li\u003e\n  \u003cli\u003eLABBench 2 (science): Argon 88.8%, Astra 85.4%, Opus 5.5 73.1%.\u003c/li\u003e\n  \u003cli\u003eVibe Code Bench: 91.9%. LVBench (long video): 91.7%. GraphWalks under 128,000 tokens: 99.7%.\u003c/li\u003e\n  \u003cli\u003eFrontierSWE v2 (coding): Argon 55.0%, below Astra\u0026#39;s 65.5% and Opus 5.5\u0026#39;s 62.3%.\u003c/li\u003e\n  \u003cli\u003eTerminal Bench 4.0: Argon about 57%, below Opus 5.5\u0026#39;s 60% and Claude Sonnet 5.5\u0026#39;s 64%. Terminal-Bench Science 0.1: 57.6% against Astra\u0026#39;s 68.1%. OSWorld-2.0: 69.2% against Astra\u0026#39;s 72.6%.\u003c/li\u003e\n  \u003cli\u003eCWE-bench v1 (vulnerability fixing): Argon 68%, tied with Astra and one point above Opus 5.5\u0026#39;s 67%. Argon ran in Antigravity, Astra in Codex, Opus 5.5 in Claude Code.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThe output cap grows from 64,000 tokens to 1 million, about eight times the 128,000-token ceiling of Opus 5.5 and Astra. Google lists four large benchmark gaps against both rivals: AutomationBench, the Harvey legal benchmark, Vals Finance Agent v2, and GraphWalks. Artificial Analysis scored Argon 77.5% on AutomationBench-AA, about six points above Claude Sonnet 5.5\u0026#39;s 71.3%.\u003c/p\u003e","blueskyPost":"Gemini 4 Argon leads the benchmarks but Google withheld it from the public, routing it only through the Fairwind Program to vetted cyber defense groups and internal staff.","twitterPost":"Argon's benchmarks lead, but Google kept it from the public. Fairwind Program access goes to vetted cyber defense groups first.","threadsPost":"Google is holding Gemini 4 Argon back. The benchmark leader is not a public release: it goes to vetted cyber defense organizations through the Fairwind Program and to thousands of Google employees, with paid API, Ultra, developer, and general access to follow in stages.","newsletterBlurb":"Google announced Gemini 4 Argon on September 30, 2026, a frontier model with a 1 million token output cap and published leads on knowledge work, long-context, and business automation benchmarks. It is not generally available. Access goes first to cyber defense organizations through the Fairwind Program, then in stages to paid API customers and Google AI Ultra users.","attributionJson":"[{\"source\":\"ASCII.jp\",\"url\":\"https://ascii.jp/elem/000/004/439/4439067/?rss\",\"title\":\"Gemini Finally Awakens? Google's \\\"Gemini 4 Argon\\\" Beats Astra and Opus 5.5 in Benchmarks\"},{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20261001-google-gemini-4-argon/\",\"title\":\"Google Announces \\\"Gemini 4 Argon,\\\" Outperforming GPT-6 Astra and Claude Fable 5.1 as the First Model in the Gemini 4 Series\"},{\"source\":\"GameBusiness.jp\",\"url\":\"https://www.gamebusiness.jp/article/2026/10/01/28193.html\",\"title\":\"Google Announces New Frontier Model Gemini 4 Argon: 1 Million Token Output, Strong at Long-Horizon Autonomous Reasoning, Large Lead Over Opus 5.5 and Astra in Legal and Business Automation\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":17707,"outputTokens":1901,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1791057749,"createdAt":"2026-10-03T19:47:00.000Z","publishedAt":"2026-10-03T19:48:25.000Z","updatedAt":"2026-10-03T19:48:25.000Z"},"cluster":{"id":"c_8c4d62532acd5fca8249d505","canonicalTitle":"Googleが「Gemini 4 Argon」を発表、GPT-6 AstraやClaude Fable 5.1を超える性能でGemini 4シリーズの先駆けモデル","representativeArticleId":"a_b12e0b1021b82821451d9239","sourceCount":3,"writtenSourceCount":3,"writeAttempts":0,"isSolo":false,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[\"Gemini 4 Argon\"],\"studios\":[\"Google\"],\"people\":[],\"type\":\"announcement\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-10-01T01:58:00.000Z","lastSeenAt":"2026-10-01T05:00:03.000Z","updatedAt":"2026-10-03T19:48:21.000Z"},"attribution":[{"source":"ASCII.jp","url":"https://ascii.jp/elem/000/004/439/4439067/?rss","title":"Gemini、ついに覚醒か　グーグル「Gemini 4 Argon」AstraやOpus 5.5超えるベンチマークも"},{"source":"GameBusiness.jp","url":"https://www.gamebusiness.jp/article/2026/10/01/28193.html","title":"Google、新フロンティアモデルGemini 4 Argon発表　出力100万トークンで長期の自律推論に強み、法務や業務自動化でOpus 5.5やAstraに大差"},{"source":"GIGAZINE","url":"https://gigazine.net/news/20261001-google-gemini-4-argon/","title":"Googleが「Gemini 4 Argon」を発表、GPT-6 AstraやClaude Fable 5.1を超える性能でGemini 4シリーズの先駆けモデル"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":["Gemini 4 Argon"],"studios":["Google"],"people":[],"type":"announcement","domain":"other","is_roundup":false},"keyFacts":["Google announced Gemini 4 Argon on September 30, 2026.","Gemini 4 Argon scored 68.9% on the Vals Index against GPT-6 Astra's 63.1% and Claude Opus 5.5's 67.0%.","Gemini 4 Argon scored 84.2% on GraphWalks in the 256,000 to 1 million token input range.","Gemini 4 Argon has an output cap of 1 million tokens, up from 64,000 tokens in the prior model.","Gemini 4 Argon introductory pricing is $2 per 1 million input tokens and $10 per 1 million output tokens, rising to $4 and $20 after the introductory period."]}