Gemini 4 Carbon

Strengths

• According to Google, it scores 59 and ranks first on the Artificial Analysis Intelligence Index v5.0, one point ahead of Claude Opus 5.5.
• Google says it leads at finding vulnerabilities (SEC-bench Pro 88.2%) and at fixing them (CWE-bench 71.0%). In Google's internal test it writes secure code more than twice as successfully as Argon.
• Against instructions hidden in documents or web pages (prompt injection), it resisted every attack in the Gray Swan test, while Argon let 0.7% through. These are Google's figures.
• Google says it can keep working for days on long engineering tasks and makes roughly half as many unsupported claims as Argon.

Weaknesses

• You cannot get it yet. Google has given no date for developers, businesses or the public, only that it will be as soon as possible.
• Google has not yet published the context window, supported inputs or other specs.
• Apart from the Artificial Analysis index, all results so far come from Google itself, and independent tests are not out yet.
• After the introductory period the price doubles to $4 for input and $20 for output.

Best for

• finding and fixing software vulnerabilities (according to Google)
• long multi-step engineering tasks (according to Google)
• cyber defense teams that get access through the Fairwind Program

Not ideal for

• everyday work today, until it is publicly available. Until then, use Gemini 4 Argon

How to use it in Europe

You cannot try it in Europe yet. Since 28 October 2026 Google has been giving it only to some cyber defenders and, step by step, to other trusted partners through the Fairwind Program. It plans to open it to developers, businesses and the public as soon as it finishes tuning the safeguards, but has given no date. Until then, Google's strongest available model is Gemini 4 Argon.

What is coming

Confirmed by vendor Google announced the model on 28 October 2026, but so far only selected cyber defenders have it. Google wants to open it to developers, businesses and consumers soon, without giving a date. When: no date yet, Google says “soon” →

Price compared with similar models

Benchmark results

Benchmarks cover only a narrow slice of ability and vendors often report them at their best settings. Treat them as a rough guide.

More models from this vendor

• Gemini 4 Argon: Google's most capable publicly available model, out in October 2026. Built for demanding analytical work where accuracy matters and made-up answers are costly.
• Gemini 3.8 Flash: Google's fast, lower-cost model for everyday work, coding and document analysis. You can try it in the API for free.
• Gemini 3.5 Flash-Lite: A very fast, low-cost Google model for large volumes of simple tasks, such as pulling data out of documents.
• Gemma 4 31B: Google's largest open model, which you can run on your own computer or server when data must not leave the company.
• Gemma 4 26B A4B: Google's faster open model for your own server. Close to the larger Gemma 4 31B in quality, but quicker to respond.

Similar models from other vendors

• Claude Opus 5.5 (Anthropic): Anthropic's main model for demanding work, coding and long documents, and the one Anthropic recommends starting with.
• DeepSeek V4 Pro (DeepSeek): DeepSeek's largest open-weight model for demanding text tasks at a low price. It does not understand images.
• Muse Wonder (Meta): Meta's strongest model for long-running agents and coding, still released as a preview.
• Mistral Large 4 (Mistral AI): Mistral's largest model, still in preview. Built for demanding tasks, long documents and security analysis with data kept in the EU.
• Mistral Large 3 (Mistral AI): Mistral's previous large model with open weights under Apache 2.0. Cheap in the API and suited to your own powerful server.
• Kimi K3.1 Pro (Moonshot AI): Moonshot AI's newest model from October 2026, built for agentic coding, office documents and very long outputs.

• Gemini pricing: what Google AI Plus, Pro and Ultra cost
• All AI prices in one place

Sources

• Google: Gemini 4 Carbon, our next frontier model ↗
• Google DeepMind: Fairwind Program ↗