Google Opens Its Most Powerful Model Gemini 4 Carbon to All
Google has officially opened its Gemini 4 Carbon model to developers, businesses, and researchers. It is Google's most powerful reasoning model so far, built for complex software engineering, multi-step agent tasks, and cybersecurity. It is now available through the Gemini API for paid developers.
The model supports text, image, video, audio, and PDF input, with a context window of over one million tokens. It costs $2 per million input tokens and $10 per million output tokens, and the deal runs until November 30, 2026.
In tests, the model scored 88.2% on the SEC-bench Pro vulnerability discovery test. Security firm Wiz once used it to find zero-day flaws in large SaaS companies in under two hours. Google positions it as a high-reasoning model for long, complex tasks, working alongside Gemini 4 Argon.
Your personal, proactive Google AI assistant that simplifies work, school and everyday life right on your phone
Fastest gateway to Gemini multimodal models with 2 M token context, caching and search grounding
Agent-first IDE with natural-language commands, tab autocomplete and synced control across editor, terminal & browser
ByteDance AI chatbot development platform with LLM integration, plugin ecosystem and cross-platform deployment to Doubao and Feishu
Comments
Loading comments…