Skip to content

GoogleFrontier

Gemini 4 Argon

First Gemini 4 model, and top of Epoch AI's GPQA Diamond runs (96.4%). Trusted cyber defenders (Fairwind) got it first; general release followed on 13 October 2026.

Announced
Released
Access
Public

Benchmark scores

TestScoreBest on the trackerHow it was measured
Epoch Capabilities Indexmany tests combined into one score167 pointsGPT-6.2 Astra 174 pointsIndependentlisted as "Gemini 4 Argon", CI 163.88 to 169.53Run 9 Nov 2026
AA Intelligence Index v5about ten tests in one score50.4 pointsClaude Fable 5.2 64.6 pointsIndependentv5.0.1; listed as Gemini 4 Argon (High)Run 10 Nov 2026
Remote Labor Indexpaid freelance projects17%Claude Fable 5.2 41%IndependentGemini Argon 4Run 22 Oct 2026
FrontierSWE V2very long engineering jobs55%Claude Fable 5.2 70%Independentproximus harness, high effort, mean@5 over 34 tasks, 20-hour budget (best@5 64.3)
MirrorCoderebuilding whole programs59%Claude Opus 5.5 79%Independenthigh reasoning effort, MirrorCode (ML, +Private, 2L): 15 programs x 2 languages, 3 runs each, 10B-token / 7-day budget, ±11.7 SERun 14 Oct 2026
APEX-Agentsprofessional office tasks82%Gemini 4 Carbon 84%Independentpass@1, unknown effort; listed as "Gemini 4 Argon"
ARC-AGI-3puzzle games, standard harnessTypical first-time human player: 100%50%Claude Fable 5.2 81%IndependentARC Prize config "Gemini 4 Argon (Low)" (google-gemini-4-argon-low), standard harness
FrontierMath T1–3hard maths problems86%Claude Fable 5.2 95%Independenthigh reasoning effort (gemini-4-argon_high)Run 13 Oct 2026
GPQA DiamondPhD-level science questionsPhD experts: 70%Saturated96%Top scoreIndependenthigh reasoning effort (gemini-4-argon_high)Run 13 Oct 2026
FrontierMath T4research-level maths problemsSaturated78%GPT-6.1 Sol 100%Independenthigh reasoning effort (gemini-4-argon_high)Run 13 Oct 2026
ARC-AGI-2visual puzzles, model aloneAverage person: 60%Saturated91%GPT-6 Astra 95%IndependentARC Prize config "Gemini 4 Argon (High)" (google-gemini-4-argon-high)
OTIS Mock AIMEcompetition mathsSaturated100%Top scoreIndependenthigh reasoning effort (gemini-4-argon_high)Run 13 Oct 2026

Compare every model on the same tests →

Also on these charts

Sources