Google Tests New-Generation Gemini 4 Carbon Version, Coding Capability Benchmarked Against Top Competitors

Google Tests New-Generation Gemini 4 Carbon Version, Coding Capability Benchmarked Against Top Competitors

AIbase

As Google prepares to release the new-generation Gemini 4 "Argon" model, internal teams have begun intensively testing a subsequent iteration codenamed "Carbon". The model has now been integrated into the internal code platform, aiming to further narrow the technological gap with top industry competitors.

Outstanding Performance on Complex Logic Tasks

According to internal documents and employee feedback, the new version Carbon performs excellently when handling code generation and complex long-term tasks, with its preliminary test level approaching the strongest agent models of its kind. As Google continues to advance in-depth testing of multiple derivative versions, the pace of technological evolution of its large model matrix is accelerating across the board.

Internal Codenames Independent of Final Naming

It is worth noting that tech companies' internal product codenames are usually not linked to the official names released publicly. For example, a previously released model under its final name actually had the internal codename "Barium-B", so it remains to be seen whether the future Carbon will debut as an independent new model or be released as a regular version.

This article comes from the AIbase Daily

Welcome to the [AI Daily] column! Here is your guide to exploring the world of artificial intelligence every day. Each day we present you with hot topics in the AI field, focusing on developers, helping you gain insight into technology trends and learn about innovative AI product applications.