AI Tools Today: Google unveils Gemini 4 Argon and HeyGen ships its first general video model

Two releases today, one from Google on the model side and one from HeyGen on the video side, and each comes with a line on what it changes for a brokerage, a property manager or an investment team.

Google unveils Gemini 4 Argon

Google announced Gemini 4 Argon today as its new frontier model, built for long, multi-step work across software engineering, enterprise knowledge work and cybersecurity. It reads and writes up to 1 million tokens, and the output limit is up from 64,000 tokens in earlier Gemini models. Google says it ranks first on the Vals Index, which measures economic impact across finance, coding, legal and tax work, and scores 51.3% on AutomationBench, a test of business function automation. (Google)

Access is limited for now to the Fairwind Program for trusted cyber defenders, and Google says it will reach developers, enterprises and consumers through the paid API and Google AI Ultra later, without giving a date. The introductory API price is $2 per million input tokens and $10 per million output tokens, with cached input priced 95% below the standard rate. (Google, Techzine)

For your office: nothing to switch this week, since most teams can’t reach it yet. When it opens up, the 1 million token output is the number to watch, because that is enough room for a full offering memorandum draft or a rebuilt rent roll to come back in one response.

HeyGen ships HeyGen Video 1

HeyGen released HeyGen Video 1, its first general purpose video model, built on MiniMax H3 and post-trained by HeyGen. Each request returns a 5 to 15 second clip at 480p or 768p with the picture and the sound generated together. You can start from a prompt, from an image that becomes the first frame, or from up to 9 reference images, 3 videos and 3 audio clips. It runs through HeyGen’s API on paid API keys. (HeyGen)

A 15 second clip is a scene, so a longer piece gets built in layers. HeyGen’s avatar videos carry the long talking segments at up to 30 minutes each, and multi-scene studio videos run as long as their scenes add up to. (HeyGen)

For your office: the reference mode takes photos of the actual building, so a listing teaser or the opening seconds of a market update starts from the real property, at a length that suits LinkedIn and Instagram. We wrote up how a 30 minute episode comes together on HeyGen and where 768p holds up.

We publish this every day there is something real to report. If a tool your office depends on shipped something we missed, tell us at realestateaistudio.com/contact.

Cover image courtesy of MiniMax, a frame from an H3 sample clip.