Key points
- Google DeepMind announced Gemini 4 Carbon, launching two weeks after Gemini 4 Argon, with access limited at first to a subset of cyber defenders through its Fairwind Program.
- Genie 4 entered research preview for trusted testers, adding native audio, free-form edits mid-simulation and scenes that hold together for up to two minutes.
- Genie 4 will reach Google AI Ultra subscribers in Project Genie and early Gemini API and AI Studio partners in the coming weeks, per TechCrunch; no API pricing yet.
- Carbon keeps Argon's introductory pricing, so teams budgeting for Argon will not see a rate change when broad access opens.
Why it matters for marketers: Marketers get a faster model cadence and a possible new storyboarding tool, but neither is broadly open yet, so planning should focus on testing and budgeting rather than rollout.
Google DeepMind announced two products in quick succession. Gemini 4 Carbon is its newest frontier model, arriving about two weeks after Gemini 4 Argon. Google says it leads the Artificial Analysis Intelligence Index and is aimed at long-running engineering work and cybersecurity defense. For now, only a subset of cyber defenders can use it, with more trusted partners added over time. Google says broader availability for developers, businesses and consumers will follow once guardrails are refined, and it is taking part in a voluntary U.S. government pre-release testing process.
Genie 4 is the company's updated world model, which generates interactive environments from prompts. According to TechCrunch, it adds audio, including music, ambient sound and dialogue, improves physics, and lets users change weather, camera angle or objects while a scene runs. Scenes stay consistent for up to two minutes. Google also acknowledges that details can drift and in-world text may not be legible.
For marketers and content teams, Genie 4 is the more interesting creative signal. TechCrunch reports that trusted testers are exploring film prototyping, and a Google DeepMind engineer pointed to ideation and storyboarding as the best fits for real-time generation. Google's engineer said the end-to-end product experience is still unclear, so this is an early experiment, not a production tool.
Watch for the Genie 4 rollout in the coming weeks to Ultra subscribers and early API partners, and for pricing, which has not been published. On the Carbon side, track when it leaves the defender-only phase. Teams already running on Argon should know Google is keeping Carbon at the same introductory rates, which simplifies any future switch.
Sources
This brief is our own summary. Read the full reporting at the source.
- Google DeepMind: Gemini 4 Carbon: Our next frontier model
- Ars Technica: DeepMind unveils Genie 4 world model promising more realistic physics and humans
- TechCrunch: Google’s Genie 4 world model adds audio, more realistic worlds, and real-time steering
- The Information: Google Launches Gemini 4 Carbon, Its Second Major AI Model in a Month
- The Decoder: Google Deepmind's Genie 4 puts a modern spin on agent designs that predate LLMs
