Artificial Intelligence
Google tests Gemini 4 Carbon against Opus 5.5
Business Insider reported that Google was testing Gemini 4 variants Argon, Barium and Carbon, with Carbon running on an internal coding platform and drawing comparisons with Anthropic's Opus 5.5.
- Business Insider said it had seen materials showing Google testing Gemini 4 variants called Argon, Barium and Carbon.
- The report said Carbon had been deployed on Google's internal coding platform Jetski over the past few days and was said to beat Argon mainly on programming tasks.
- One employee compared Carbon with Anthropic's Opus 5.5, while another said early Argon versions reminded them of Opus 5 on some coding tasks.
- Google described Argon as its frontier reasoning model, and the article said the company mapped its models to different roles.
- The piece said the Gemini app showed a new Automatic mode and adjustable reasoning intensity, while Google AI Studio surfaced an Ultra mode.
Business Insider said it had seen internal documents, screenshots and chats showing Google testing Gemini 4 variants named Argon, Barium and Carbon. Carbon had been run on Google's internal coding platform Jetski over the past few days and was said to outperform Argon mainly on programming tasks.
One employee compared Carbon with Anthropic's Opus 5.5, although that person said more testing was still needed. Another employee said early Argon versions brought to mind Opus 5 on some coding work, even though internal feedback was positive overall.
Google described Argon as its frontier reasoning model and said its models were assigned to different roles, including reasoning, speed, generative media and lightweight edge workloads. The article said Carbon and Barium were likely checkpoints or updates inside the Argon family, and that Argon had previously been called Barium-B.
The piece said Google had not announced a Gemini 4 launch date. It also said the Gemini app now showed an Automatic mode for 3-series models and adjustable reasoning intensity from low to high, while Google AI Studio had surfaced an Ultra mode.
Named in this story
People
- Vedant Misra
- responded on X to the Business Insider report
- Logan Kilpatrick
- said the team was working on getting the most out of Argon
Companies
- was testing Gemini 4 variants and reworking its apps
- Anthropic
- was the company behind Opus 5.5 and Opus 5
Organisations
- Business Insider
- was cited as having seen internal materials on the models
- Google DeepMind
- employed Vedant Misra
Products and systems
- Gemini 4
- was the model line under discussion
- Argon
- was described by Google as its frontier reasoning model
- Barium
- was one of the internal Gemini 4 variants
- Carbon
- was another internal Gemini 4 variant tested on Jetski
- Jetski
- was Google's internal coding platform where Carbon was deployed
- Opus 5.5
- was the Anthropic model one employee compared Carbon with
- Opus 5
- was the older Anthropic model that reminded one employee of early Argon
- Gemini app
- showed new modes and adjustable reasoning controls
- Google AI Studio
- surfaced an Ultra mode
How the source tells it
The register was speculative and mildly promotional, leaning on internal claims and model comparisons while stressing imminent launch activity.
- speculation presented as fact internal documents and employee impressions were used to stand in for confirmed performance
- hype high-end model labels and self-improvement language framed the rollout as a leap forward
- urgency repeated emphasis on already and launch preparation created a sense of imminent change