Gemini 4 Carbon
Positioning and release
Gemini 4 Carbon is a frontier reasoning model based on Gemini 4 Argon, designed for complex, long-running agentic workflows, software engineering, scientific reasoning, and cybersecurity. Google announced it on October 28, 2026, initially offering it to a subset of cybersecurity defenders. On November 10, 2026, Google announced broader availability to paying developers and enterprises, with rollout to Google AI Ultra subscribers across selected consumer products. (blog.google)
Capabilities and uses
- Accepts text, images, audio, and video and produces text. Google Cloud documents a combined input-and-output limit of 1,048,576 tokens; extended output depends on the API and continuation mechanism used. (deepmind.google)
- Intended for multi-step coding and debugging, cybersecurity defense, vulnerability discovery and patching, document and data analysis, and scientific research. Thinking levels can be configured to balance quality, latency, and cost. (deepmind.google)
- Google reported scores of 88.2% on SEC-bench Pro, 54.8% on ExploitGym, and 71.0% on CWE-bench. Google states that these cybersecurity benchmark evaluations were conducted without guardrails; results should not be treated as guarantees in production. (blog.google)
Limitations and access
The model can hallucinate, may be slow or time out, and can consume more tokens at higher thinking levels. Its stated knowledge cutoff is March 2026, although knowledge in some domains may be older. Google continues to improve resistance to jailbreaks; outputs and security-sensitive actions require appropriate human review. Specialized defensive-cyber capabilities are available to verified defenders through the Fairwind Program. (deepmind.google)
Pricing
Gemini API, Google AI Studio paid tier: prices below are in USD per 1 million tokens, as checked on November 16, 2026. There is no free tier for this model. Output pricing includes thinking tokens; only implicit context caching is supported. (aistudio.google.com)
| Service tier | Through November 30, 2026: input / output / cached input | From December 1, 2026: input / output / cached input |
|---|---|---|
| Standard | $2.00 / $10.00 / $0.10 | $4.00 / $20.00 / $0.20 |
| Batch | $1.00 / $5.00 / $0.05 | $2.00 / $10.00 / $0.10 |
| Flex | $1.00 / $5.00 / $0.05 | $2.00 / $10.00 / $0.10 |
| Priority | $3.60 / $18.00 / $0.18 | $7.20 / $36.00 / $0.36 |
These are Google AI Studio Gemini API rates, not a statement of prices on every Google Cloud platform or in every region. Grounding and other services may incur separate charges. (aistudio.google.com)
Official sources
- Original model announcement — October 28, 2026. (blog.google)
- Availability announcement — November 10, 2026. (blog.google)
- Google DeepMind model card — capabilities and limitations. (deepmind.google)
- Google AI Studio pricing — pricing checked November 16, 2026. (aistudio.google.com)