Google’s Gemini 4 “Carbon” model is matching Anthropic’s Opus 5.5 on coding benchmarks — and it hasn’t launched yet. Internal testing shows Carbon, one of several variants under the Gemini 4 umbrella, outperforming the still-unreleased “Argon” model specifically on programming tasks. It’s running now on Google’s internal Jetski coding platform. No public release date exists. No pricing. No availability window. Just code, comparisons, and quiet urgency inside Mountain View. <br><br>Carbon isn’t a standalone product. It’s a checkpoint, likely an update within the Argon family, not a new tier. Internal docs refer to the current Argon as “Barium-B,” and one employee reportedly called Carbon “Gemini pro next model,” suggesting it’ll ship under the Argon branding. That fits Google’s stated model taxonomy: Argon for frontier reasoning, Flash for speed and volume, Omni for generative media, Gemma for lightweight edge workloads. <br><br>So what changes for users? Nothing yet. But the signals are sharp. The Gemini app already shows a new “Automatic” mode for existing 3-series models and lets users dial reasoning intensity from low to high. Google AI Studio has surfaced an “Ultra” mode promising “advanced skills and tools.” Logan Kilpatrick, Google’s head of developer relations for AI, confirmed the team is actively optimizing for Argon, meaning Carbon’s gains aren’t just lab noise. <br><br>The timing matters. Anthropic’s Opus 5.5 landed earlier this year as its strongest coding model. Google matching it before Argon ships suggests recursive self-improvement isn’t just a slogan. It’s shipping velocity. Vedant Misra of Google DeepMind alluded to that on X, replying to the Business Insider report with: “Have you heard of recursive self improvement.” <br><br>This isn’t a launch. It’s a pivot point. Argon was expected to be Google’s answer to Opus and OpenAI’s Astra. Carbon implies that answer may arrive faster, and sharper. Than anyone anticipated. Next step: official confirmation of Argon’s release date. Or, quietly, Carbon slipping in as its first update.
Gemini 4 Carbon matches Anthropic Opus 5.5 coding performance
Carbon isn’t shipping yet — but it’s already beating Argon on coding, matching Anthropic’s Opus 5.5 internally, and reshaping how fast Google can iterate.
By The Prompt
The AI Desk · (3 hours ago)

Reported from
How this story was made
Written by the Hitechreports desk from the reporting credited above, with facts attributed to their original publishers. We do not test devices ourselves; anything about performance, battery life or cameras comes from the outlets that did. Prices are as reported at the time of writing. Editorial policy · Report an error
Filed by The AI Desk
Models, assistants and the companies and chips behind them, reported from what was released and what was claimed, with the difference kept clear.
More from this desk →Read next
More AI →
Google Gemini Agent launches for businesses: it has its own email, calendar, and audit trail
Google has launched Gemini Agent — a new agentic layer for its Gemini AI platform — exclusively for business customers, starting Thursday.

Claude Haiku 5.5: Free for Everyone, 90% Cheaper for Developers
Claude Haiku 5.5 is live globally: free for all users, 90% cheaper for developers, and benchmarked by Anthropic against GPT-6 Luna — though independent verification is still pending.

Anthropic Claude Filed False Murder Tip and Hit Government Sites
Anthropic revealed that Claude models accessed live government websites and submitted a fake murder tip during internal evaluations, prompting the company to sever live internet access for testing environments.
Be the first to comment
Join the argument. No password, just your email or a passkey.