By the toolist.ai editors. Researched from public sources, not hands-on tested by toolist.ai. Updated November 3, 2026; prices checked November 3, 2026.
Gemini 4 Carbon: price, context window and benchmarks
Gemini 4 Carbon is the model Google calls its strongest, announced on October 28, 2026. You cannot buy it yet: it is going first to a subset of cyber defenders and then to trusted partners, and Google gives no date for everyone else. Google says it will launch at Gemini 4 Argon's introductory $2 per million input tokens and $10 per million output tokens, then cost $4 and $20 after an introductory period with no published end date. Artificial Analysis, measuring it before general release, puts it at 60 on its Intelligence Index at high thinking, the highest score of any model on this site.
Best for
- A subset of cyber defenders, the first group Google is giving access to, followed by trusted partners in its Fairwind Program
- Long engineering work once it opens: 65% on Terminal-Bench 4.0 at high thinking, against 60% for Claude Opus 5.5
- Finding and fixing software flaws once it opens: Google reports 88.2% on SEC-bench Pro and 71.0% on CWE-bench, both evaluated without guardrails
Key facts
| Fact | Gemini 4 Carbon | Source |
|---|---|---|
| Provider | Google: Gemini 4 Carbon: our next frontier model | |
| API model name | Not published yet | Google: Gemini 4 Carbon: our next frontier model |
| Announced | October 28, 2026 | Google: Gemini 4 Carbon: our next frontier model |
| Input price | $2 per million tokens | Google: Gemini 4 Carbon: our next frontier model |
| Cached input price | $0.10 per million tokens | Google: Gemini 4 Carbon: our next frontier model |
| Output price | $10 per million tokens | Google: Gemini 4 Carbon: our next frontier model |
| Context window | 1,000,000 tokens | Artificial Analysis: Gemini 4 Carbon (high) |
| Input and output | Text, images, speech and video in; text out | Artificial Analysis: Gemini 4 Carbon (high) |
| Reasoning | Artificial Analysis tests low, medium and high thinking. Google has not published a default level | Artificial Analysis: Gemini 4 Carbon (high) |
| API and cloud platforms | Not generally available. A subset of cyber defenders first, then additional trusted partners in phases through the Fairwind Program. Google says developers, businesses and consumers will follow as soon as possible and gives no date | Google: Gemini 4 Carbon: our next frontier model |
Announced prices; Carbon is not on sale yet, and Google does not say whether its first groups pay. Google says Carbon will launch at the same introductory price as Gemini 4 Argon, with cached input 95% below the input price, then $4 per million input tokens and $20 per million output tokens. Argon's introductory period ends November 30, 2026; Google does not say whether Carbon's ends then too. Its Gemini API pricing page does not list Carbon yet, and Artificial Analysis lists $4 and $20. The 1,000,000-token context window is Artificial Analysis's rounded figure; Google has not published an exact limit.
What Gemini 4 Carbon costs a month
Three typical workloads. These costs use the introductory price Google has announced for Gemini 4 Carbon. Nobody outside Google's first access groups can use Carbon yet, and the price rises to $4 and $20 per million tokens after an introductory period with no published end date.
| Workload | Tokens a month | Cost a month |
|---|---|---|
| Support chatbot 100,000 replies, each 3,000 input and 400 output tokens, no caching | 300M in, 40M out | $1,000 |
| Document Q&A 5,000 questions over the same 60,000-token document set (90% cache hits), 800 output tokens each | 300M in, 4M out | $127 |
| Coding agent 200 tasks, each 1.5 million input tokens (90% cache hits) and 50,000 output tokens | 300M in, 10M out | $187 |
Reasoning models bill their thinking as output tokens, so heavy reasoning can raise the output count several times. Cache-write charges are not included.
Benchmarks
| Benchmark | Score | Source |
|---|---|---|
| Artificial Analysis Intelligence Index high thinking, the highest level Artificial Analysis tests | 60.0 | Artificial Analysis: Gemini 4 Carbon (high) |
| Intelligence Index at everyday effort medium thinking; Google has not published a default | 58.0 | Artificial Analysis: Gemini 4 Carbon (medium) |
| Terminal-Bench 4.0 high thinking | 65.0% | Artificial Analysis: Gemini 4 Carbon vs Claude Opus 5.5 |
| HLE-Diamond high thinking | 69.0% | Artificial Analysis: Gemini 4 Carbon vs Claude Opus 5.5 |
| Terminal-Bench-Science 0.1 high thinking | 65.0% | Artificial Analysis: Gemini 4 Carbon vs Claude Opus 5.5 |
Speed: about 88.7 output tokens a second (high thinking, Artificial Analysis; Artificial Analysis: Gemini 4 Carbon (high)).
Benchmark and speed data marked as such come from Artificial Analysis.
What changed
Google builds Carbon on Gemini 4 Argon, which launched two weeks earlier. The figures in this paragraph are Google's own. Without guardrails it reports 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploiting them and 71.0% on CWE-bench for patching them. It reports a gain of 89 Elo points, to 1768 and fourth place, on WebDev Arena. Attack success on Gray Swan's Indirect Prompt Injection Benchmark falls from Argon's 0.7% to 0.0%, and successful cheating on Terminal-Bench 4.0 is 0.0%. Google's post quotes 59 on version 5.0 of the Artificial Analysis Intelligence Index. (Google: Gemini 4 Carbon: our next frontier model)
Gemini 4 Carbon compared
Other models
- GPT-6.1 Terra: $1 in, $5 out per million tokens
- Gemini 4 Argon: $2 in, $10 out per million tokens
- Muse Wonder: $4 in, $20 out per million tokens
- Claude Haiku 5.5: $0.10 in, $0.50 out per million tokens
- GPT-6.1 Sol: $2 in, $10 out per million tokens
- Claude Sonnet 5.5: $2 in, $10 out per million tokens
- Claude Opus 5.5: $4 in, $20 out per million tokens
- GPT-6 Luna: $0.10 in, $0.50 out per million tokens
- GPT-6 Sol: $2 in, $10 out per million tokens
- GPT-6 Astra: $10 in, $50 out per million tokens
- Claude Fable 5.1: $10 in, $50 out per million tokens
- GLM-5.3: $1.40 in, $4.40 out per million tokens
- Claude Opus 5: $5 in, $25 out per million tokens
All prices side by side: AI model API price table. How we research: our method.
Sources
- Google: Gemini 4 Carbon: our next frontier model (checked 2026-11-03)
- Gemini API docs: Pricing (checked 2026-11-03)
- Artificial Analysis: Gemini 4 Carbon (high) (checked 2026-11-03)
- Artificial Analysis: Gemini 4 Carbon (medium) (checked 2026-11-03)
- Artificial Analysis: Gemini 4 Carbon vs Claude Opus 5.5 (checked 2026-11-03)