← Sunday, November 15 deck

10Product

Varun Mohan Says Antigravity Doubled Context for Argon, Carbon and Hydrogen

We’re doubling the context window for Argon, Carbon, and Hydrogen in Antigravity. Lots of work to make this happen but you should now see compaction go down by 3.5x per user turn!

Context

Why it matters

A larger context window means coding agents can keep more of a long task in view and spend less time compressing earlier work, which can make long jobs finish faster.

Varun Mohan, who builds Antigravity at Google DeepMind, posted on the evening of November 15 that the product is doubling the context window for Argon, Carbon and Hydrogen. He said users should now see compaction fall by 3.5 times per user turn. The post does not give the old or new token counts, say which plans are included, or explain how the 3.5 times figure was measured.

Deep dive

Mohan's post is the official notice of the change. Antigravity is Google's agent coding environment, and Argon, Carbon and Hydrogen are Gemini 4 models offered inside it. Earlier official posts from Google presented Carbon as the stronger Gemini 4 model and Hydrogen as a faster model in the same family, with Argon already in the lineup. This update is about the context window those models get in Antigravity, not a new model release.

Compaction is the step a long chat takes when earlier turns are summarized so the session can stay inside the model's context limit. Mohan said doubling the window should cut that compression by 3.5 times on each user turn. He called the work substantial, but the post does not define the baseline, the tasks used for the measurement, or the new window size in tokens. No matching note had been added to Antigravity's public blog at the time this file was prepared.

Because the post is short, several practical details remain open, including whether every Antigravity plan gets the new window, whether quota use rises when more tokens stay in context, and whether the 3.5 times drop holds on large coding jobs as well as short chats. The claim should be read as Mohan's product statement until Google publishes measured figures.

What to expect next

Developers will be able to tell quickly whether long agent sessions in Antigravity compact less often and finish sooner. Watch for official token limits, plan eligibility, and any quota guidance from Google or from Mohan. If the larger window is limited to higher plans, the practical impact will be narrower than the announcement implies. A later note on the Antigravity blog would also help confirm the change in writing.