Google shipped three different kinds of AI system in October 2026, and the naming does not make the differences obvious. Gemini 4 Argon is a reasoning model for hard knowledge work. Gemini 4 Carbon is a longer-running specialist for engineering and cybersecurity. Genie 4 is neither: a world model that generates interactive, sounding simulations in real time. This guide explains what each is for, how they differ, and what a business can realistically do with each.
Argon is the general-purpose frontier model
Gemini 4 Argon launched on October 13, 2026, and rolled out globally through the Gemini API, to Google AI Ultra subscribers in Antigravity, and to Ultra and Pro subscribers in the Gemini app and AI Mode in Search. Google positions it as a frontier reasoning model for complex, multi-step work across coding, finance, legal, cybersecurity, and science, and it ranked first on the Vals Index, which measures performance across sectors such as finance, coding, legal, and tax work.
The engineering results show what that means in practice. The Chrome team used Argon to separate core webpage handling from a roughly six-million-line legacy folder inside Chromium's 40-million-line codebase, finishing a refactor across hundreds of files in weeks rather than months. Researchers made the quantum chemistry package xTB up to 2.3 times faster and pushed the open Gemma 4 E2B model to a 2.85 times inference speedup on an A100 GPU with no quality loss. In robotics, Argon scored 87% on the Embodied Reasoning benchmark and, in an internal test, watched a cornhole video, built the simulation code itself, and transferred the result to Apptronik's Apollo humanoid, which sank seven consecutive shots in a real trial.
Argon also powers the Fairwind Program, which gives vetted cyber defenders access without cyber guardrails, and Google is offering 10,000 scientists six months of free access to Antigravity's top tier.
API pricing is published: $2 per million input tokens and $10 per million output tokens, with cached input at 95% off. A token is roughly a word fragment, so the price is best read as a few dollars for a substantial document analysis. The introductory rate holds until November 30, 2026; from December 1 the price doubles to $4 and $20.
Carbon is built for long, security-critical runs
Google announced Gemini 4 Carbon on October 28, two weeks after Argon, and the gap between the announcements is the clearest signal that these are different tools. Carbon is designed for long-horizon engineering and cybersecurity defense: Google says internal teams have run it for days to root-cause and patch subtle bugs across large codebases and low-level systems. It ranks first on the v5.0 Artificial Analysis Intelligence Index with a score of 59.
The cybersecurity numbers are the headline. Carbon scores 88.2% on SEC-bench Pro for finding real vulnerabilities, 54.8% on ExploitGym for validating that they are exploitable, and 71.0% on Collinear's CWE-bench for patching root causes. It generates secure code at more than twice Argon's success rate on internal benchmarks, and the Chrome team reported more than ten times better performance than another lab's large frontier model on high-quality patches. Wiz also found it leading on web penetration testing benchmarks.
Carbon also tightens alignment. Successful cheating on Terminal-Bench 4.0 fell to 0.0%, high-severity misalignment alerts dropped almost 80%, unsupported claims were cut roughly in half, and it resisted every evaluated prompt-injection attack on the Gray Swan benchmark, where Argon had a 0.7% success rate.
Where Argon is the broad frontier model, Carbon is the specialist: harder, longer tasks, a stronger safety stack, and a deliberate rollout. It began in pre-release testing with the U.S. government and cyber defenders through the Fairwind Program; a broader launch to developers, businesses, and consumers is still to come. Pricing matches Argon: $2 and $10 per million input and output tokens, doubling to $4 and $20 after the introductory period.
Genie 4 is a world model, not a chat model
Genie 4, announced on October 27, 2026, is a different category of system. A world model learns how an environment behaves and then predicts what happens next, frame by frame, in response to your actions. Instead of generating one fixed video clip, it renders a world you can move through and change. Genie 4 is natively multimodal: it generates the visuals and the sound together, including ambient noise, music, and interactive dialogue in which characters react to what you do and say.
You steer it with keyboard controls and natural-language prompts. Ask for a change in weather, a new object, or a different camera angle, and Gemini adapts the request to the scene and keeps the world coherent between inputs. Genie 4 simulates physics such as aerodynamics and collisions, and its most dramatic leap is human movement: characters move naturally, respond to their surroundings, and can be directed in real time. Other Gemini models supply the reasoning that keeps the simulation coherent. For complex tasks, Gemini 3.8 Flash acts as a guiding agent, watching the frames, breaking a goal into steps, and checking that each step succeeded.
Genie 4 is a research preview for trusted testers, with access first to Google AI Ultra subscribers in Project Genie and early access partners in the Gemini API and Google AI Studio. Through the API it generates continuous 720p worlds at 24 frames per second with synchronized audio. Pricing is $1.50 per million input tokens for text or reference images, $9 per million output tokens for text, and $17.50 per million output tokens for video, audio included. That works out to 240 tokens per frame, or about $0.10 per second, roughly $6 a minute.
How the three fit together
The relationship is layered rather than competitive. Gemini models handle reasoning, planning, and language; Genie renders the physical world those plans move through. Google demonstrated the combination by pairing Argon with Genie 4 and Gemini Robotics: Argon breaks a user's instruction into steps and feeds them to Genie 4 as text captions while the simulation renders, so the world adapts as the task progresses. In one demonstration, the pair simulated Apptronik's Apollo humanoid playing checkers. Argon also has a dedicated robotics variant, and Genie 4 is the training ground where robots and agents practice before touching real hardware.
What a business can do with each
Argon fits professional knowledge work: contract review, financial analysis, codebase refactoring, and scientific literature synthesis. Enterprises can reach it through Gemini for business, the API, and Antigravity, and scientists can apply for the free six-month Antigravity tier. Carbon fits security operations and long-running engineering agents, especially vulnerability discovery, validation, and patching, though broad access is still ramping up and defenders can apply through Fairwind. Genie 4 fits creative and training pipelines: film pre-visualization, archival photographs turned into explorable scenes, game prototyping before engine code, and varied environments for robotics and agent training. All three carry SynthID watermarks where media is generated, so output can be identified as AI-made.
If you are deciding which model deserves a pilot project, or how world simulation could fit your AI plans, Crowe Endeavors tracks these releases and what they mean for practical deployment.
Sources
"Google AI updates October 2026 (Gemini 4 Argon, Carbon, Genie 4)" (Google, October 2026) — https://www.youtube.com/watch?v=sFfPkUJd_Jk
"The latest AI news we announced in October 2026" (Google blog, October 2026) — https://blog.google/innovation-and-ai/technology/ai/google-ai-updates-october-2026/
"Genie 4: A new generation of world simulation" (Google DeepMind, October 27, 2026) — https://deepmind.google/blog/genie-4-a-new-generation-of-world-simulation/