Google has unveiled Gemini 4 Carbon, a new frontier model it describes as its strongest yet for cybersecurity and long-horizon work. The model follows Gemini 4 Argon by only two weeks and will initially reach a subset of cyber defenders this week. Google plans to add trusted partners gradually through its Fairwind Program, then bring Carbon to developers, businesses and consumers through its products after collecting feedback and iterating on guardrails.
I did say we'd be releasing our models periodically...introducing Gemini 4 Carbon, our next frontier model.
— Sundar Pichai (@sundarpichai) October 28, 2026
Following Argon's release just two weeks ago, Carbon shows the pace at which we’re pushing the frontier, it ranks #1 on the recently updated v5.0 of @ArtificialAnlys… pic.twitter.com/GN9LqZ7Zcf
Carbon ranks first on version 5.0 of the Artificial Analysis Intelligence Index with a score of 59. In external cybersecurity evaluations conducted without guardrails, it scored 88.2% on SEC-bench Pro for vulnerability discovery, 54.8% on ExploitGym for exploitation and 71.0% on Collinear’s CWE-bench for producing working fixes. Google says internal tests also showed more than twice Argon’s success rate for secure-code generation, while the Chrome team found it produced high-quality patches at more than 10 times the rate of another lab’s large frontier model.
Google’s Gemini 4 Carbon is the new #1 on the Artificial Analysis Intelligence Index v5, and second only to Claude Opus 5.5 on the Artificial Analysis Coding Agent Index
— Artificial Analysis (@ArtificialAnlys) October 28, 2026
Gemini 4 Carbon scores 59 on Intelligence Index v5, driven by leading performance in complex reasoning,… pic.twitter.com/gSH65FSKzL
The model is built for sustained agentic workflows across coding, science, full-stack web development and multimodal understanding. Google’s teams have seen Carbon run for days, work through dead ends and complete multi-step engineering tasks across large codebases and low-level systems. Its web development performance rose by 89 Elo points to fourth place on WebDev Arena with a score of 1769.
Exciting news: Gemini 4 Carbon (High) from @GoogleDeepMind just made a huge debut on Arena, landing #1 in Text Arena, #2 in Agent Arena, and #4 in Code Arena: WebDev. Back at the top, Gemini 4 Carbon delivers frontier intelligence and reshapes the Pareto frontier across three… https://t.co/gryD3OJGS9 pic.twitter.com/I8Rlg3SsbX
— Arena.ai (@arena) October 28, 2026
Google DeepMind is pairing those capabilities with the safety and security stack introduced for Argon, spanning model training, real-time misuse defenses, prompt-injection resistance and chain-of-thought monitoring. Carbon recorded 0.0% successful cheating on Terminal-Bench 4.0, with attempts at 0.4%, while high-severity misalignment alerts in internal coding work fell by almost 80%. Google also reports roughly half as many unsupported claims, 26% less overreach and 29% less under-service than Argon. Its attack success rate on Gray Swan’s Indirect Prompt Injection Benchmark fell from Argon’s 0.7% to 0.0%.
Carbon will launch at Argon’s introductory price of $2 per million input tokens and $10 per million output tokens. Cached inputs will cost 95% less than standard input tokens. The staged release reflects Google’s participation in the US government’s voluntary pre-release testing process, with wider access held back while the company tests safeguards with trusted cyber defenders.
Sources and related context
- Gemini 4 Argon: Built to solve your hardest problems: Related context: Google's earlier Argon release explains the model Carbon builds on and the introductory pricing it is expected to share.
- Fairwind Program: Related context: Google DeepMind describes the trusted defender access program through which Carbon's initial partner rollout is planned.
