Google is making its newest AI model, Gemini 4 Carbon, available to all paying developers and enterprises on Nov 10 2026. The rollout follows a limited test with “cybersecurity defenders” two weeks earlier. Carbon is marketed as Google’s most capable model yet, aimed at tasks that push the limits of AI, such as long‑chain coding and open‑ended scientific research.
At the same time, Google is previewing a “swarm” capability on its Antigravity platform. The feature lets a single user prompt spawn a coordinated team of AI agents that work together on a problem. The company says this is designed for long‑horizon projects that need multiple specialized steps.
How Carbon fits into Google’s model lineup
Google launched a smaller frontier model called Argon last month. Argon remains the recommended choice for most software‑engineering and knowledge‑work tasks. Carbon, by contrast, is positioned for more demanding use cases that require higher reasoning depth. The company says Carbon “pushes the frontier even further” than Argon.
Pricing for Carbon starts at $2 per million input tokens and $10 per million output tokens during an introductory period. That is roughly one‑fifth the cost of OpenAI’s GPT‑6 Astra and one‑quarter the cost of Anthropic’s newly released Fable 5.2. After the introductory period ends at the end of November, the rates will double but will still be cheaper per token than the competitors’ frontier models.
What the benchmarks show
Google cites internal benchmark results that place Carbon among the top models on most metrics, second only to Anthropic’s Fable 5.2. The company notes that benchmark numbers can be selective, so independent testing is still needed to confirm the claims.
Why it matters
Developers and businesses can now access a more powerful Google model at a lower per‑token cost, making advanced AI tasks like multi‑step coding cheaper than using rival services. The swarm preview may let users build more complex workflows without writing extensive glue code, though the feature is still in preview and not yet widely available.
