How much does the Gemini 4 Carbon API cost?

Introductory API pricing is $2.00 per 1M input tokens, $10.00 per 1M output tokens, and $0.10 per 1M cached input tokens. Post-introductory rates double for inputs and outputs.

What is the context window of Gemini 4 Carbon?

Gemini 4 Carbon features a 1,048,576-token context window and a maximum output generation of 65,536 tokens.

Gemini 4 Carbon: Pricing, Benchmarks & Access

Q: What is Gemini 4 Carbon?

Gemini 4 Carbon is an advanced AI model announced by Google DeepMind on October 28, 2026, targeted at cybersecurity defense and long-horizon engineering.

Q: How can I access Gemini 4 Carbon?

Public access is currently restricted to select cyber defenders and trusted partners through the Fairwind Program.

Gemini 4 Carbon API Price and Benchmarks. A precision-milled aluminum server chassis sits open under a harsh amber strobe light, revealing dense arrays of liquid-cooled silicon wafers crisscrossed by heavy black power cabling — GPUniq

Data as of October 29, 20263 sources

Drafted with AI tools and edited by a human. The figures and conclusions were checked by a GPUniq editor.

TL;DR

Google DeepMind announced Gemini 4 Carbon on October 28, 2026. Introductory API pricing sits at $2.00 per 1M input tokens and $10.00 per 1M output tokens, with cached inputs priced at $0.10 per 1M tokens. The model features a 1,048,576-token context window and a 65,536-token maximum output. According to the Google DeepMind Official Blog, Carbon scores 88.2% on SEC-bench Pro and targets cybersecurity defense and long-horizon engineering. Access is restricted to select cyber defenders and trusted partners via the Fairwind Program.

API Pricing

Google DeepMind structured pricing into introductory and post-introductory tiers. Input tokens cost $2.00 per 1M tokens initially. Output tokens run $10.00 per 1M tokens. Post-introductory rates double these figures to $4.00 and $20.00 per 1M tokens.

Token Type	Introductory Price, $/1M tokens	Post-Introductory Price, $/1M tokens
Input	2.00	4.00
Output	10.00	20.00
Cached Input	0.10	Not Published

Context Window and Limits

The model accepts a 1,048,576-token context window. Maximum output generation peaks at 65,536 tokens. Google DeepMind has not yet published detailed technical specifications or a specific model ID at launch.

Benchmark Performance

Carbon improves over its predecessor, Gemini 4 Argon. According to the Google DeepMind Official Blog, Carbon hits 59 on the Artificial Analysis Intelligence Index v5.0 while Argon scores 50.

Model	Intelligence Index v5.0	SEC-bench Pro, %	ExploitGym, %	CWE-bench, %
Gemini 4 Carbon	59	88.2	54.8	71.0
Gemini 4 Argon	50	83.4	42.0	68.0

Cybersecurity benchmarks show strong gains in vulnerability detection and remediation. According to the Google DeepMind Official Blog, Carbon achieves 88.2% on SEC-bench Pro compared to Argon at roughly 83.4%. Exploit validation on ExploitGym reaches 54.8% against Argon's 42%. CWE-bench patching scores land at 71.0% versus Argon's 68%.

Availability

Public access is currently restricted. According to the Google DeepMind Official Blog, the model is available to a subset of cyber defenders. Google is expanding access gradually through trusted partners via the Fairwind Program. Broader availability is not yet scheduled, and weights or architectural details remain unreleased.

Methodology

Facts are compiled from the vendor's own announcement and documentation, linked below, at the time of writing; where the vendor hasn't published a number it is marked as such. GPUniq prices, when the model is already available here, come from our live catalog.