{"slug": "gemini-4-argon-google-s-flagship-model-shares-a-name-with-a-gas-that-barely", "title": "Gemini 4 Argon: Google's Flagship Model Shares a Name With a Gas That Barely Reacts", "summary": "Google released Gemini 4 Argon, its newest frontier model, on 13 October 2026 at $2 per 1M input tokens and $10 per 1M output tokens through 30 November 2026, with a 1M-token context window and a March 2026 knowledge cutoff. Google's launch post said the model \"is now globally available today for developers and enterprises,\" but the Gemini API pricing page states Gemini 4 Argon \"is not available on the Free tier or on the Google AI Pro plan,\" restricting it to paid API tiers and Google AI Ultra subscribers. Google said it is \"looking forward to bringing you the rest of the Gemini 4 family\" without naming models or dates.", "body_md": "# Gemini 4 Argon: Google's Flagship Model Shares a Name With a Gas That Barely Reacts\n\nGemini 4 Argon went on general release on 13 Oct 2026: $2 input and $10 output per 1M tokens through 30 Nov, a 1M token window, and where you can use it.\n\nOur view: Gemini 4 Argon is worth a test on your longest agent jobs, but the launch price is a countdown. Google put Gemini 4 Argon on general release on 13 October 2026 at $2 input and $10 output per 1M tokens through 30 November 2026. Checked 13 Oct 2026.\n\nNew models drive coding agents. [Munder Difflin](https://harnessmd.com/download) is free and open source: a desktop app that runs a team of coding agents such as Claude Code, Codex and Gemini CLI on your own computer. Gemini 4 Argon is already in the app’s model pickers for Antigravity and Cursor, added by a [pull request](https://github.com/HarnessMD/munder-difflin/pull/738) merged on 13 Oct 2026. The [install guide](https://munderdiffl.in/blog/how-to-install-and-use-munder-difflin/) covers setup, and the [Concepts hub](https://munderdiffl.in/blog/topics/concepts/) explains the terms.\n\n## What is Gemini 4 Argon? [#](#what-is-gemini-4-argon)\n\nGemini 4 Argon is Google’s newest top model, built for long jobs in coding, office work and security. Google’s [developer post](https://blog.google/innovation-and-ai/technology/developers-tools/build-with-gemini-4-argon/) of 13 October 2026 calls it “our frontier model for complex workflows across real-world software engineering, enterprise knowledge work, and cybersecurity defense”.\n\nGoogle’s posts do not say why it is called Argon. The element is a noble gas that reacts with almost nothing, which is not the sales pitch here.\n\nThe rows come from Google’s [model card](https://deepmind.google/models/model-cards/gemini-4-argon/) and [models page](https://ai.google.dev/gemini-api/docs/models), read on 13 Oct 2026.\n\n|  | Gemini 4 Argon | \n|---|---|\n| API model ID | gemini-4-argon | \n| Input | Text, images, audio and video | \n| Output | Text | \n| Context window | Up to 1M tokens, input and output combined | \n| Knowledge cutoff | March 2026 | \n\nIt is the first of a set. Google’s [launch post](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/) says “We’re looking forward to bringing you the rest of the Gemini 4 family”, with no names and no dates. The [Hacker News thread](https://news.ycombinator.com/item?id=50073879) had more than 400 points on 13 Oct 2026.\n\n## When was Gemini 4 Argon released? [#](#when-was-gemini-4-argon-released)\n\nGemini 4 Argon went on general release on 13 October 2026. Google’s launch post, dated Oct 13, 2026, says the model “is now globally available today for developers and enterprises”.\n\nIt was not a surprise. [The Verge](https://www.theverge.com/tech/1002980/google-gemini-4-argon) reported on 30 September 2026 that Google had revealed the model and was holding it for “trusted cyber defenders” at first. The curtain opened in two pulls.\n\n## Where can you use Gemini 4 Argon? [#](#where-can-you-use-gemini-4-argon)\n\nYou can use it through the paid Gemini API, in Google AI Studio and Antigravity, and in the Gemini app, depending on what you pay for. The rows are our summary of Google’s launch post and its [Antigravity post](https://antigravity.google/blog/gemini-4-argon-in-google-antigravity), both dated 13 Oct 2026.\n\n| Where | Who gets it | \n|---|---|\n| Gemini API | Paid tier | \n| Google AI Studio | Paid tier, or Google AI Ultra subscribers | \n| Android Studio | Paid tier of the Gemini API | \n| Antigravity desktop app and CLI | Google AI Ultra subscribers | \n| Gemini app and AI Mode in Search | Google AI Ultra and Pro subscribers | \n\nThe free door is shut. Google’s [pricing page](https://ai.google.dev/gemini-api/docs/pricing) says Gemini 4 Argon “is not available on the Free tier or on the Google AI Pro plan”. That note sits on the API pricing page; the launch post still lists Pro subscribers for the Gemini app and AI Mode in Search, and Google’s [plans page](https://one.google.com/about/google-ai-plans/) says it is “rolling out to Google AI Pro subscribers in select countries”. For businesses, the Antigravity post lists the API in Gemini platform and Antigravity “on eligible Gemini Enterprise subscriptions”.\n\n## What does Gemini 4 Argon cost? [#](#what-does-gemini-4-argon-cost)\n\nGemini 4 Argon costs $2.00 input and $10.00 output per 1M tokens through 30 November 2026, then $4.00 and $20.00 from 1 December 2026, on Google’s pricing page as read on 13 Oct 2026. The output price includes thinking tokens.\n\n| Paid tier, per 1M tokens | Through 30 Nov 2026 | From 1 Dec 2026 | \n|---|---|---|\n| Input | $2.00 | $4.00 | \n| Output | $10.00 | $20.00 | \n| Context caching | $0.10 | $0.20 | \n\nIntroductory prices are the free trial of the API world: pleasant until the calendar turns. Budget on the December column. The pricing page lists batch prices at half of each figure.\n\n## How does Gemini 4 Argon score against Claude Opus 5.5 and GPT-6 Astra? [#](#how-does-gemini-4-argon-score-against-claude-opus-55-and-gpt-6-astra)\n\nOn Google’s own table, Gemini 4 Argon has the top score in 13 of 19 rows, shares one and trails in five. The four coding rows split two and two. These are Google’s figures from its [Gemini 4 Argon page](https://deepmind.google/models/gemini/), read on 13 Oct 2026.\n\n| Coding benchmark | Gemini 4 Argon | GPT-6 Astra | Claude Opus 5.5 | \n|---|---|---|---|\n| DeepSWE v1.1 | 77.9% | 74.1% | 74.2% | \n| Vibe Code Bench | 91.9% | 89.6% | 90.3% | \n| FrontierSWE v2 | 55.0% | 65.5% | 62.3% | \n| Terminal-bench 4.0 | 57.4% | 58.2% | 66.4% | \n\nA vendor table that prints rows it loses is rarer than it should be. Read the bottom two before you switch anything.\n\n## How does Google say Gemini 4 Argon behaves as an agent? [#](#how-does-google-say-gemini-4-argon-behaves-as-an-agent)\n\nGoogle reports fewer shortcuts and better resistance to planted instructions, and leaves one security question open. The launch post says that on “coding benchmarks where models have no incentive to hold back, shortcut attempts dropped to just 2%”, and that attack success on the Gray Swan Indirect Prompt Injection benchmark fell to 0.7% “across 15 adaptive attempts”.\n\nThe model card is blunter about one risk: “We cannot rule out that Gemini 4 Argon reaches the cybersecurity CCL”. CCL is a risk threshold the model card tests against. The card then says “We are therefore applying safeguards that we assess lower risk to below the cybersecurity CCL level”.\n\n## Should you switch your coding agents to Gemini 4 Argon? [#](#should-you-switch-your-coding-agents-to-gemini-4-argon)\n\nTest it on long jobs first, and do not move everything at once. Our reasons: it tops most of Google’s table, the coding rows are split, and the price doubles on 1 December 2026.\n\nMore on this launch: [Gemini 4 Argon vs Claude Opus 5.5 vs GPT-6 Astra](https://munderdiffl.in/blog/gemini-4-argon-vs-claude-opus-5-5/), [Gemini 4 Argon pricing](https://munderdiffl.in/blog/gemini-4-argon-pricing/), [how to use it](https://munderdiffl.in/blog/how-to-use-gemini-4-argon/) and [what it did inside Google](https://munderdiffl.in/blog/gemini-4-argon-inside-google/).\n\nFor how it sits beside Google’s other tools, see [Gemini CLI vs Claude Code](https://munderdiffl.in/blog/gemini-cli-vs-claude-code/) and [Claude Code vs Antigravity](https://munderdiffl.in/blog/claude-code-vs-antigravity/). For Meta’s rival launch this week, see [Muse Wonder](https://munderdiffl.in/blog/muse-wonder/).\n\n## FAQ\n\nWhen was Gemini 4 Argon released?\n\nGemini 4 Argon went on general release on 13 October 2026. Google's launch post of that date says it is now globally available for developers and enterprises. The Verge reported the first reveal on 30 September 2026, when access was limited to a set of trusted cyber defenders.\n\nHow much does Gemini 4 Argon cost?\n\nGoogle's pricing page, read on 13 Oct 2026, lists Gemini 4 Argon at $2.00 input and $10.00 output per 1M tokens through 30 November 2026, then $4.00 input and $20.00 output from 1 December 2026. Output includes thinking tokens. Batch prices are half of each figure.\n\nIs Gemini 4 Argon free?\n\nNo. Google's pricing page says Gemini 4 Argon is not available on the Free tier or on the Google AI Pro plan. Google's launch post lists the paid tier of the Gemini API, Google AI Ultra subscribers in AI Studio and Antigravity, and Google AI Ultra and Pro subscribers in the Gemini app and AI Mode in Search.\n\nWhat is the Gemini 4 Argon context window?\n\nGoogle's model card lists a combined input and output context window of up to 1M tokens. Google's models page describes a shared 1M context window and long decodes up to 1M tokens. The API model ID is gemini-4-argon.\n\nIs Gemini 4 Argon better than Claude Opus 5.5 at coding?\n\nIt depends on the test, on the figures Google published. Google's table has Gemini 4 Argon ahead on DeepSWE v1.1 (77.9% against 74.2%) and Vibe Code Bench (91.9% against 90.3%), and Claude Opus 5.5 ahead on Terminal-bench 4.0 (66.4% against 57.4%) and FrontierSWE v2 (62.3% against 55.0%).", "url": "https://wpnews.pro/news/gemini-4-argon-google-s-flagship-model-shares-a-name-with-a-gas-that-barely", "canonical_source": "https://munderdiffl.in/blog/gemini-4-argon/", "published_at": "2026-10-13 00:00:00+00:00", "updated_at": "2026-10-16 22:25:50.163369+00:00", "lang": "en", "topics": ["large-language-models", "generative-ai", "ai-products", "artificial-intelligence"], "entities": ["Google", "Gemini 4 Argon", "Google AI Studio", "Antigravity", "Gemini API", "Google AI Ultra", "Munder Difflin", "The Verge"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/gemini-4-argon-google-s-flagship-model-shares-a-name-with-a-gas-that-barely", "markdown": "https://wpnews.pro/news/gemini-4-argon-google-s-flagship-model-shares-a-name-with-a-gas-that-barely.md", "text": "https://wpnews.pro/news/gemini-4-argon-google-s-flagship-model-shares-a-name-with-a-gas-that-barely.txt", "jsonld": "https://wpnews.pro/news/gemini-4-argon-google-s-flagship-model-shares-a-name-with-a-gas-that-barely.jsonld"}}