AI Trend Research

← October 10, 2026 briefing · added in the 12:00 KST update

P2 Model Unconfirmed Language modelsAgents12:00 updateGoogleAnthropic

Google is testing 'Carbon,' an internal checkpoint following Gemini 4 'Argon,' reportedly at "Opus 5.5 level" (report)

Announced October 10, 2026

What happened

According to a Business Insider report (10th, local time) cited by AI Times, Google employees are testing 'Carbon,' a new internal checkpoint of Gemini 4, on 'Jetski,' an internal-only coding platform. Carbon is reportedly a version with improved coding performance. Google announced Argon on September 30, giving some cybersecurity partners early access, and also built Argon into the enterprise Gemini agent it unveiled on the 8th.

Provider claims

The "Opus 5.5 level" in the headline is an assessment contained in the report; it has not been confirmed by an official Google announcement or public benchmarks.

Why it matters

The report says a follow-up checkpoint with stronger coding performance has appeared before Argon is even generally available. It can be read as a sign that Google is trying to close the gap with Anthropic's top model in the coding model race.

Confidence low · official source pending

Sources

More about Google