Google's most powerful model Gemini 4 Carbon now widely available to paying developers and Ultra users

Google announced on November 10 that Gemini 4 Carbon, its strongest model, is now widely available to paying developers, enterprises and Google AI Ultra subscribers. Carbon is built for multi-step coding, cross-system debugging and long-horizon scientific reasoning, and Google calls it its most powerful model to date.

Availability and pricing

Developers can use Carbon through the paid tier of the Gemini API in AI Studio and Android Studio; enterprises can access it via Gemini for Business, Antigravity and CodeMender; Ultra subscribers can try it in Antigravity, the Gemini app and AI Mode in Search. The introductory price matches Argon: $2 per million input tokens and $10 per million output tokens. Google is also previewing /swarm in Antigravity, which turns a single prompt into a coordinated team of agents handling roles such as architectural design, coordination and auditing.

Benchmark results

Google says Carbon ranks first on three independent benchmarks — CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench — at more than twice lower cost per task than the next best external model. On the new Multimodal Knowledge Agent (MKA) benchmark, developed with Artificial Analysis, Carbon ranks second while costing more than six times less per task on average than the top model. It also performed at the top in specialist evaluations such as deep financial research, ranking first out of 23 frontier model configurations on Balyasny Asset Management's General Financial Intelligence benchmark.

Cybersecurity and scientific reasoning

Carbon is also Google's strongest cybersecurity model to date: offered to defenders through the Fairwind Program, it leads on benchmarks for finding, exploiting and patching vulnerabilities. In Google's Project Fortify system, Carbon achieved 100% recall on scanning and 95% precision on triage; Google's Cloud Vulnerability Research team used it to prove that two bugs in a popular open-source project were exploitable vulnerabilities, comparable in severity to the recent PHP2Shell exploit. In science, researchers used Carbon to prove key security properties of the code underlying Signal's cryptographic protocol, and to advance homolog retrieval for protein structure prediction; drug toxicity firm Axiom Bio has moved its production scientific analyses to Carbon.

Safety measures strengthened in step

Google says safety protections were strengthened alongside Carbon's stronger cyber capabilities: in simulated cyberattack testing, Carbon blocked 99% of attack trajectories, cutting the per-attempt breach rate to 0.8%; on Gray Swan's indirect prompt injection benchmark, the attack success rate was 0%. On a 375-task agentic honesty benchmark, Carbon passed 92% of tasks, up from 70% for Argon, declining to fill in missing experimental data 98% of the time.

How it differs from Argon

Notably, Carbon does not replace Argon: Google says Argon remains the recommended choice for most software engineering and knowledge work, while Carbon is aimed at tasks that test the limits of a model's intelligence, such as complex multi-step coding, cross-subsystem troubleshooting and open-ended scientific research. Carbon was first released to cybersecurity defenders two weeks ago; this is the wider release to developers and consumers. Google opened applications four weeks ago to give 10,000 scientists free access to Argon for six months, and Carbon is now included in that program.