Physix Frontier · News Briefing Card (enbrief · Oct 2, 2026)

Google Releases Gemini 4 Argon, Tops Benchmarks But Limited to Security Teams

KEY FACTS

Google released its new flagship model Gemini 4 Argon, which ranks first on multiple evaluation leaderboards.
Argon scored 77.9% on DeepSWE v1.1, surpassing Opus 5.5 and GPT-6 Astra.
The model is currently only available to vetted cybersecurity teams, with general users unable to access it for now.
Argon supports up to one million token output and targets complex workflows in programming, finance, and law.
Some Google employees are skeptical of the model's actual coding performance, believing front-end design remains a weak spot.

KEY DATA

77.9%DeepSWE v1.1 Score

68%CWE-bench v1 Score

68.90%Vals Index Score

2 USDPrice per Million Input Tokens

PHYSIX OBSERVATION

Google is using low pricing and long context to claw its way back to the table, but handing the first batch only to security teams shows it still has concerns about abuse risk. Leading on benchmarks does not equal leading in experience, and internal doubts about the front-end weak spot are worth noting. For developers, the cost-performance ratio is a genuine positive, but whether it delivers still depends on real-world testing once it is fully opened up.

Source: enbrief original report ↗