ModelsPublished Top storyAll news from that day
Ongoing case: Frontier model API prices Price change · 15 stories
Gemini 4 Carbon opens to some cyber defenders first, with no developer date
Google calls Gemini 4 Carbon its strongest model; it costs $2 and $10 per million tokens at first, then $4 and $20, and the benchmark scores are Google's own.
Check our sources · 16 facts from 1 source
Image: Google
Key points
- Google says Gemini 4 Carbon, announced 28 October 2026, goes this week to a subset of cyber defenders, with more trusted partners phased in through its Fairwind Program.
- Google gives no date for developers, businesses and consumers, saying only "as soon as possible" while it gathers feedback and iterates on guardrails.
- The introductory price is $2 per million input tokens and $10 per million output tokens; Google says $4 and $20 apply once the introductory period ends.
What happened
Google announced Gemini 4 Carbon on 28 October 2026 and calls it its strongest model, with state-of-the-art cybersecurity and long-horizon capabilities. Carbon is not generally available yet.
Google says it is making Carbon available this week to a subset of cyber defenders, and will phase in more trusted partners through its Fairwind Program. It says developers, businesses and consumers will follow as soon as possible, and gives no date.
Google also says it is actively engaged in the US government's voluntary process for pre-release model testing. The introductory price is $2 per million input tokens and $10 per million output tokens, with cached input tokens 95% cheaper; Google says $4 and $20 per million will apply once the introductory period ends.
The benchmark results in the post are Google's own.
Google says Carbon ranks first on version 5.0 of the Artificial Analysis Intelligence Index with a score of 59, and reports 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation and 71.0% on Collinear's CWE-bench for patching, evaluated without guardrails.
On safety, Google says Carbon reduced successful cheating on Terminal-Bench 4.0 to 0.0% and cut high-severity misalignment alerts in its own live coding by almost 80%. Carbon follows Gemini 4 Argon, which Google says launched two weeks earlier.
What it means for you
Our viewMost readers cannot use Gemini 4 Carbon yet, so no product changes today. Google says wider access depends on feedback and guardrail work, and it gives no date. The listed price is introductory, so a budget built on $2 and $10 per million tokens should also be run at $4 and $20.
The results in the post, including first place on the Artificial Analysis Intelligence Index at 59, are Google's own statements, and Google says its external benchmark results were evaluated without guardrails.
If you work in security, ask Google whether your organisation can join the Fairwind Program; if you build on Gemini, cost your workload at both prices before committing.
It adds no new facts.
Your reaction
We count reactions per story and day, never who reacted. The counts help us choose what goes in the monthly issue. If you are signed in, your own page shows yours too.
We count reactions per story and day, never who reacted. They help choose what goes in the monthly issue.
Check our sources
Every sentence above is checked against this source.
1 Gemini 4 Carbon: our next frontier model
Open the source archived copy-
Google published "Gemini 4 Carbon: our next frontier model", by Koray Kavukcuoglu, SVP, Google DeepMind and Chief AI Architect, on 28 October 2026.
citeGemini 4 Carbon: our next frontier model. Oct 28, 2026. Koray Kavukcuoglu, SVP, Google DeepMind and Chief AI Architect, Google
-
Google says: "Today, we’re announcing the next frontier model: Gemini 4 Carbon, our strongest model with state-of-the-art cybersecurity and long horizon capabilities."
citeToday, we’re announcing the next frontier model: Gemini 4 Carbon, our strongest model with state-of-the-art cybersecurity and long horizon capabilities.
-
Google says: "We're taking a deliberate approach to rollout and are actively engaged in the U.S. government’s voluntary process for pre-release model testing."
citeWe're taking a deliberate approach to rollout and are actively engaged in the U.S. government’s voluntary process for pre-release model testing.
-
Google says: "Starting this week, we will make Carbon available to a subset of cyber defenders and will gradually phase access to additional trusted partners through our Fairwind Program."
citeStarting this week, we will make Carbon available to a subset of cyber defenders and will gradually phase access to additional trusted partners through our Fairwind Program.
-
Google says it is "continuing to gather feedback and iterate on guardrails before making Carbon available to developers, businesses, and consumers via our product surfaces as soon as possible", and the post gives no date.
citeWe’re continuing to gather feedback and iterate on guardrails before making Carbon available to developers, businesses, and consumers via our product surfaces as soon as possible.
-
Google says Gemini 4 Carbon "will launch at the same introductory price as Argon at $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off input token price."
citeGemini 4 Carbon will launch at the same introductory price as Argon at $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off input token price.
-
Google's footnote says: "After the introductory period expires, the price of $4 per 1M input tokens and $20 per 1M output tokens will apply."
citeAfter the introductory period expires, the price of $4 per 1M input tokens and $20 per 1M output tokens will apply.
-
Google says: "Carbon ranks #1 on the recently updated v5.0 of Artificial Analysis Intelligence Index at 59." This is Google's statement of the index result.
citeCarbon ranks #1 on the recently updated v5.0 of Artificial Analysis Intelligence Index at 59.
-
Google reports Carbon "landing a firm lead in SEC-bench Pro (88.2%) for finding vulnerabilities, ExploitGym (54.8%) for exploitation, and CWE-bench (71.0%) for patching."
citelanding a firm lead in SEC-bench Pro (88.2%) for finding vulnerabilities, ExploitGym (54.8%) for exploitation, and CWE-bench (71.0%) for patching.
-
Google says its external benchmark results were "evaluated without guardrails".
citeCarbon excels across key external benchmarks (evaluated without guardrails):
-
Google says Carbon "jumps +89 elo points to land at #4 (1768) on WebDev Arena."
citeCarbon jumps +89 elo points to land at #4 (1768) on WebDev Arena.
-
Google says: "the Chrome team has reported that Carbon performed more than 10X better than another lab's large frontier model when generating high quality patches." The post does not name the other model.
citethe Chrome team has reported that Carbon performed more than 10X better than another lab's large frontier model when generating high quality patches.
-
Google says that on internal benchmarks "Carbon achieves more than 2x uplift in success rate for generating secure code over Argon."
citeOn internal benchmarks, Carbon achieves more than 2x uplift in success rate for generating secure code over Argon.
-
Google says: "Carbon reduced successful cheating on Terminal-Bench 4.0 to 0.0%, with attempts to cheat dropping to 0.4%", and that in live coding environments with its internal teams "high-severity misalignment alerts dropped by almost 80%."
citeCarbon reduced successful cheating on Terminal-Bench 4.0 to 0.0%, with attempts to cheat dropping to 0.4% — all while improving overall accuracy. In live coding environments with our internal teams, we observed that high-severity misalignment alerts dropped by almost 80%.
-
Google says that on the Gray Swan Indirect Prompt Injection Benchmark Argon had a 0.7% attack success rate and "Carbon achieves 0.0%".
citeWhile Argon previously led the Gray Swan’s Indirect Prompt Injection (IPI) Benchmark in prompt injection resistance with a very low 0.7% attack success rate, Carbon achieves 0.0% — resisting every evaluated attack attempt.
-
Google says Gemini 4 Carbon "builds on the same multi-layered safety and security stack of Argon", and describes Argon as "launched two weeks ago".
citeGemini 4 Carbon builds on the same multi-layered safety and security stack of Argon ... Building on Gemini 4 Argon, which was launched two weeks ago
Every sentence above is checked against this source. Each fact has its own address you can cite.
Source 1 · Google · 28 Oct 2026
Gemini 4 Carbon: our next frontier model
Google published "Gemini 4 Carbon: our next frontier model", by Koray Kavukcuoglu, SVP, Google DeepMind and Chief AI Architect, on 28 October 2026.
Gemini 4 Carbon: our next frontier model. Oct 28, 2026. Koray Kavukcuoglu, SVP, Google DeepMind and Chief AI Architect, Google
Google says: "Today, we’re announcing the next frontier model: Gemini 4 Carbon, our strongest model with state-of-the-art cybersecurity and long horizon capabilities."
Today, we’re announcing the next frontier model: Gemini 4 Carbon, our strongest model with state-of-the-art cybersecurity and long horizon capabilities.
Google says: "We're taking a deliberate approach to rollout and are actively engaged in the U.S. government’s voluntary process for pre-release model testing."
We're taking a deliberate approach to rollout and are actively engaged in the U.S. government’s voluntary process for pre-release model testing.
Show all 16 factsShow fewer facts
Google says: "Starting this week, we will make Carbon available to a subset of cyber defenders and will gradually phase access to additional trusted partners through our Fairwind Program."
Starting this week, we will make Carbon available to a subset of cyber defenders and will gradually phase access to additional trusted partners through our Fairwind Program.
Google says it is "continuing to gather feedback and iterate on guardrails before making Carbon available to developers, businesses, and consumers via our product surfaces as soon as possible", and the post gives no date.
We’re continuing to gather feedback and iterate on guardrails before making Carbon available to developers, businesses, and consumers via our product surfaces as soon as possible.
Google says Gemini 4 Carbon "will launch at the same introductory price as Argon at $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off input token price."
Gemini 4 Carbon will launch at the same introductory price as Argon at $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off input token price.
Google's footnote says: "After the introductory period expires, the price of $4 per 1M input tokens and $20 per 1M output tokens will apply."
After the introductory period expires, the price of $4 per 1M input tokens and $20 per 1M output tokens will apply.
Google says: "Carbon ranks #1 on the recently updated v5.0 of Artificial Analysis Intelligence Index at 59." This is Google's statement of the index result.
Carbon ranks #1 on the recently updated v5.0 of Artificial Analysis Intelligence Index at 59.
Google reports Carbon "landing a firm lead in SEC-bench Pro (88.2%) for finding vulnerabilities, ExploitGym (54.8%) for exploitation, and CWE-bench (71.0%) for patching."
landing a firm lead in SEC-bench Pro (88.2%) for finding vulnerabilities, ExploitGym (54.8%) for exploitation, and CWE-bench (71.0%) for patching.
Google says its external benchmark results were "evaluated without guardrails".
Carbon excels across key external benchmarks (evaluated without guardrails):
Google says Carbon "jumps +89 elo points to land at #4 (1768) on WebDev Arena."
Carbon jumps +89 elo points to land at #4 (1768) on WebDev Arena.
Google says: "the Chrome team has reported that Carbon performed more than 10X better than another lab's large frontier model when generating high quality patches." The post does not name the other model.
the Chrome team has reported that Carbon performed more than 10X better than another lab's large frontier model when generating high quality patches.
Google says that on internal benchmarks "Carbon achieves more than 2x uplift in success rate for generating secure code over Argon."
On internal benchmarks, Carbon achieves more than 2x uplift in success rate for generating secure code over Argon.
Google says: "Carbon reduced successful cheating on Terminal-Bench 4.0 to 0.0%, with attempts to cheat dropping to 0.4%", and that in live coding environments with its internal teams "high-severity misalignment alerts dropped by almost 80%."
Carbon reduced successful cheating on Terminal-Bench 4.0 to 0.0%, with attempts to cheat dropping to 0.4% — all while improving overall accuracy. In live coding environments with our internal teams, we observed that high-severity misalignment alerts dropped by almost 80%.
Google says that on the Gray Swan Indirect Prompt Injection Benchmark Argon had a 0.7% attack success rate and "Carbon achieves 0.0%".
While Argon previously led the Gray Swan’s Indirect Prompt Injection (IPI) Benchmark in prompt injection resistance with a very low 0.7% attack success rate, Carbon achieves 0.0% — resisting every evaluated attack attempt.
Google says Gemini 4 Carbon "builds on the same multi-layered safety and security stack of Argon", and describes Argon as "launched two weeks ago".
Gemini 4 Carbon builds on the same multi-layered safety and security stack of Argon ... Building on Gemini 4 Argon, which was launched two weeks ago
Topics
The morning email
On the mornings we publish: the three top stories and up to four short ones. Free.
We email you a link to confirm. An issue may include one sponsor, always labelled Sponsored · Advertisement. Our emails count opens and clicks, not who made them. Unsubscribe in one click. What we keep