Google DeepMind releases Gemini 4 Carbon, positioning it as its strongest model for complex reasoning and agentic workflows
Gemini 4 Carbon is now available to developers, enterprises, scientists, and consumers, with claims of frontier performance in cybersecurity, long-horizon agentic workflows, and improved coding and scientific reasoning capabilities.
1 source · cross-referenced
- Gemini 4 Carbon is Google DeepMind’s newest frontier model, available to developers, enterprises, scientists, and consumers.
- The model is positioned as the strongest in Google DeepMind’s lineup, excelling in cybersecurity, long-horizon agentic workflows, complex coding, and scientific reasoning.
- Gemini 4 Carbon ranks at the top on benchmarks such as CursorBench 4.0, Perplexity’s WANDR, and Zapier’s AutomationBench, with claims of over 2X lower cost per task compared to the next best external model.
- Pricing is set at $2 per million input tokens and $10 per million output tokens, matching the introductory price of Gemini 4 Argon.
- The model is available via API in Google AI Studio, Gemini for business, and to Google AI Ultra subscribers in Google Antigravity, the Gemini app, and AI Mode in Search.
Google DeepMind announced the availability of Gemini 4 Carbon, positioning it as the company’s strongest model to date. The release targets developers, enterprises, scientists, and consumers, with a focus on tasks that demand deep reasoning, such as complex coding, debugging, and open-ended scientific research.
The model is described as achieving frontier performance in cybersecurity capabilities and long-horizon agentic workflows, areas where extended reasoning and verification are critical. Google DeepMind states that Carbon pushes the frontier further than its predecessor, Gemini 4 Argon, which remains the preferred choice for general software engineering and knowledge work.
Gemini 4 Carbon ranks at the top on independent benchmarks designed to test real-world workflows, including CursorBench 4.0, Perplexity’s WANDR, and Zapier’s AutomationBench. The company claims Carbon delivers these results at more than 2X lower cost per task compared to the next best external model.
Pricing for Gemini 4 Carbon is set at $2 per million input tokens and $10 per million output tokens, matching the introductory price of Gemini 4 Argon. The model is available via the API in Google AI Studio and Gemini for business for paying developers and enterprises. It is also rolling out to Google AI Ultra subscribers in Google Antigravity, the Gemini app, and AI Mode in Search.
For complex projects, Google DeepMind is previewing a feature called /swarm in Google Antigravity, which turns a single prompt into a coordinated team of agents. This allows developers to assign Carbon to roles requiring deep reasoning, such as architectural design and auditing, while pairing it with lighter models for parallel tasks and adversarial verification.
In higher-precision knowledge work, such as legal, finance, and enterprise operations, Gemini 4 Carbon combines frontier visual understanding, reasoning, and exhaustive retrieval to synthesize insights across dense documents and data. Google DeepMind notes that existing benchmarks do not adequately capture performance in agentic workflows involving dense files, charts, images, and tables, and claims Carbon ranks highly on a new Multimodal Knowledge Agent benchmark developed in partnership with Artificial Analysis.
The model’s capabilities are demonstrated through examples such as turning historical photos and engineering plates into an interactive 3D exhibition of the Eiffel Tower’s construction using a single prompt, and building an interactive 906K-triangle reconstruction of Google’s Platform 37 office from seven public City Council planning PDFs. Additionally, Carbon matched the performance of the best model tested on deep financial research in pre-release evaluations, including Hebbia’s E2E Financial Agentic evaluation and Balyasny Asset Management’s General Financial Intelligence benchmark.
- Nov 11, 2026 · Google AI — Blog
Google launches Gemini Decisions API with new Gemini Decision 1 model for fast, calibrated multimodal decisions
Trust78 - Nov 11, 2026 · OpenAI — News
OpenAI offers free ChatGPT Plus to more than 20,000 veterans and servicemembers
Trust75 - Nov 11, 2026 · Latent Space — swyx
Anthropic releases Claude Fable 5.2 and Mythos 5.2, claiming top LLM spot on Anthropic’s IPO roadshow day
Trust76