Traders raise bets but stay split on Google's public Gemini 4 release by mid-October

Traders lifted expectations for Google to make Gemini 4 Argon publicly available by mid-October, but the outcome remains close to even. Bets rose over a two-hour stretch on Oct. 3, reversing part of an earlier decline. No new announcement explains the rise, and the reason for this move is not yet clear.

Expectations have swung sharply since Oct. 2. Bets climbed that Friday and rose further early Saturday, then fell before rebounding. The latest rise followed a previous rebound that had brought expectations back near even. For the nearer Oct. 10 window, a public release is now unlikely; by Oct. 31, traders consider it all but certain. That contrast suggests expectations center on a release later in October, rather than by mid-month.

Google announced Gemini 4 Argon on Sept. 30, saying it would initially be available only to selected cybersecurity testers through its Fairwind program. That restricted access does not amount to a public release. In its Oct. 2 roundup, Google said testing with trusted cyber defenders was continuing before access expands to developers, businesses and consumers. Reuters reported that Google gave no timeline for the broader rollout. Earlier reporting also described employee concerns about Argon's performance on practical tasks, particularly coding; Google disputed that account, and Axios reported Google's assertion that the model performed strongly in internal use. None of this reporting establishes what prompted the latest change in bets.

Google says it plans to gather feedback and strengthen safeguards before expanding access, with paid API customers and Google AI Ultra subscribers among the groups slated to receive access earlier. No public-release date has been announced. A Google announcement opening access to the general public by mid-October would meet the market's outcome; continued access limited to trusted testers or other selected groups would not. Google has announced no dated event bearing on the question.