MODEL

Gemini 4 Argon

modeltopic-notegoogle

Overview

Gemini 4 Argon is Google‘s Oct 1, 2026 frontier-model release — raising the output ceiling to 1M tokens (from 64k on prior Gemini generations) to support multi-step agent workflows and threat-hunting chains without API splicing. Introductory pricing lands at $2/$10 per Mtok with a 95% cached-input discount — identical on the per-token line to Claude Sonnet 5.5 (Sep 28) and GPT-6.1 Sol (Sep 29). The public pricing page lists a post-introductory standard rate of $4/$20 — double the launch rate — with no published end date for the intro window. Access is gated to Google’s Fairwind cyber-defender program first, with paid API customers and AI Ultra subscribers next.

Timeline

  • 2026-10-01-AI-Digest — Google ships Gemini 4 Argon at $2/$10 per Mtok introductory (standard rate posts at $4/$20) with a 1M-token output ceiling and Fairwind-gated cyber-defender early access. Closes the convergent mid-tier cluster around Claude Sonnet 5.5, GPT-6.1 Sol, and Grok 4.7 within one quarter. Load-bearing softener: the public pricing page lists a $4/$20 standard rate — the $2/$10 is a teaser that will roll off, so any downstream app-cost model should budget against the standard. Separately, early community framing of Argon as “cyber-defense-first” is overstated — Argon’s own positioning covers coding, enterprise knowledge work, and cyber defense, and the headline architectural change is a general-capability 1M-token output window. Fairwind is a distribution gate (early access with relaxed cyber guardrails for vetted defenders), not a specialized model fork. Reframe worth carrying: Google ships a general frontier model with intro pricing matching the Sep 28–29 mid-tier cluster, under Fairwind-gated cyber-defender early access, not Google ships a specialized cyber-defense model at Sonnet 5.5 pricing. Log against MOC - Major Companies and MOC - Open Source Models.

Key Developments

  1. Fourth Frontier Lab Onto the $2/$10 Mid-Tier Line Within One Quarter (October 1, 2026): Argon’s $2/$10 intro pricing lands on the same per-token line as Claude Sonnet 5.5 (Sep 28) and GPT-6.1 Sol (Sep 29), joining Grok 4.7 ($2/$6, Sep 21) in a convergent mid-tier cluster — four labs on the same cost band inside one quarter. Load-bearing framing: convergent competitive repricing, not coordination — and the standard rate on Argon is $4/$20, so the floor may not hold through 2027.

  2. 1M-Token Output Window — First Frontier Model Lifting Output Ceiling From 64k (October 1, 2026): Argon raises the output ceiling to 1M tokens from the prior 64k Gemini default — the headline architectural change, designed to support multi-step agent workflows and threat-hunting chains without API splicing. The structural read is that output-length ceilings are now the binding constraint on long-horizon agent workflows, not context-window length.

  3. Fairwind-Gated Early Access, Not a Specialized Cyber Fork (October 1, 2026): Access is gated to Google’s Fairwind cyber-defender program first, with paid API customers and AI Ultra subscribers next. Framing to carry: Fairwind is a distribution gate (early access with relaxed cyber guardrails for vetted defenders), not a specialized model variant — Argon’s own positioning covers coding, enterprise knowledge work, and cyber defense as a general-capability model.

  • 2026-10-02-AI-Digest — Gemini 4 Argon surfaces as (1) comparator anchor in Cloudflare‘s Clef launch — Clef sits in front of LLMs like Claude Opus 5.5 or Argon on the control-plane layer — and (2) as the second explicit mid-tier-step-up data point in today’s Key Takeaway on the convergent $2/$10 floor. Load-bearing framing to carry: Argon’s $4/$20 standard rate (the post-introductory posted price) joins Claude Opus 5.5‘s step-up as the second data point suggesting the convergent-mid-tier $2/$10 line will not hold through 2027 as a standard rate — intro periods will. Only two labs so far have published explicit step-up curves, but the direction is consistent. Reframe: signal, not confirmation — two labs published step-up curves out of four on the mid-tier line, not the mid-tier floor is dead. Log against MOC - Major Companies and MOC - Open Source Models.

  • 2026-10-14-AI-Digest — Google opened Gemini 4 Argon worldwide on 2026-10-13 via API / Gemini app / Antigravity, ending the Oct 1 Fairwind-gated early-access window (already-reported: 2026-10-01-AI-Digest). Benchmark reads: Google’s own table shows Argon winning 12 of 18 benchmarks vs Claude Opus 5.5 (DeepSWE 77.9 vs 74.2, Harvey Legal 19.6 vs 3.8); Vals Index places Argon #1 at 68.90% vs Claude Opus 5.5 at 67.0%. Claude Opus 5.5 still leads on Terminal-bench 4.0 (66.4 vs 57.4) and FrontierSWE v2 (62.3 vs 55.0). Aider polyglot top-5 (fetched 2026-10-14) carries no Argon entry — standings unchanged, so the practitioner reference signal hasn’t rotated yet despite the worldwide landing. Load-bearing softener: The Decoder’s “trails Claude Opus 5.5 and Meta Muse Wonder Full in independent benchmarks” framing is category-mixed on the Meta side — Muse Wonder is Meta’s consumer-agent frontier model, not a reasoning benchmark apex competitor. Reframe worth carrying: Argon trades benchmark wins with Opus 5.5, does not trail it, not Argon worldwide landing trails Opus 5.5 and Muse Wonder. Watch: a non-Google / non-Vals third-party benchmark pass before promoting Argon-leads out of softener territory. Log against MOC - Major Companies.

See also: Google, Gemini, Claude Sonnet 5.5, GPT-6.1 Sol, Grok 4.7, MOC - Major Companies, MOC - Open Source Models.

  • 2026-10-22-AI-Digest — Wiz used Gemini 4 Argon to independently report the same critical PHP RCE that OpenAI credits GPT-5.6 Sol for discovering under the Oct 21 Patch the Planet batch disclosure. The PHP Foundation shipped a fix on Oct 21. This is the first publicly recorded AI-assisted CVE confirmed across two labs’ research pipelines on different frontier models. Load-bearing framing: AI-assisted CVE discovery is a running series now reaching higher-blast-radius software, not first AI-discovered CVE. Argon’s role here is as Wiz’s AI-assisted research tool rather than a product beat. Log against MOC - Agent Security.

  • 2026-10-29-AI-Digest — Argon surfaces today as the predecessor comparator in Google DeepMind‘s Oct 28 launch of Gemini 4 Carbon — ~two weeks after Argon went GA worldwide (Oct 13 per 2026-10-14-AI-Digest). Per The Decoder, Carbon posts AAII v5 at 59 vs Argon’s 50 — a nine-point index jump — with Carbon still trailing Claude Opus 5.5 on agentic coding. Pricing holds at $2/$10 per Mtok (same intro-tier line Argon introduced on Oct 1). Load-bearing softener: the Argon→Carbon cadence looks fast on paper, but Google has been iterating Gemini 4 variants on Arena since September — the first-order claim is Carbon ships with the autonomous-engineering + prompt-injection angle, not the frontier release cadence is accelerating. Aider polyglot top-5 (fetched 2026-10-29) still carries no Gemini 4 entry — standings unchanged. Log against MOC - Major Companies.