Every major AI-news aggregator spent the first half of July reporting that Gemini 3.5 Pro — Google’s flagship reasoning model, teased at I/O back in May — would ship on 17 July with a 2-million-token context window and pricing around $15/$60 per million input/output tokens. We referenced the same date in our 13 July signal on the strength of that reporting. It’s now 19 July, and Google’s own Gemini API documentation still lists Gemini 3.5 Flash as the newest stable model in the family. There’s no Gemini 3.5 Pro entry, no model card, and no pricing page. Google has never issued a blog post, spokesperson quote, or developer-console listing confirming a 17 July date — tracing the story back far enough shows it originated with a couple of tech blogs citing unnamed sources, then got repeated by dozens of others until it read as established fact.
That gap between “widely reported” and “officially confirmed” matters more than it looks. Gemini 3.5 Pro almost certainly exists — Google confirmed as much on stage in May, and the model has reportedly slipped once already after an internal rebuild. But a date with no primary source attached is a rumour wearing a release note’s clothes, and it’s exactly the kind of claim that’s easy to repeat without checking.
So what
If a stakeholder is holding off on an AI tooling decision — which model to build against, which vendor to pick, whether to wait for “the next big release” — because of a date they saw in a roundup post, it’s worth two minutes to check whether that date traces back to the vendor itself or to a chain of blogs quoting each other. We apply the same standard to every AI-tooling claim that ends up in a client recommendation on an AI-assisted development engagement: verified against a primary source, or flagged as unconfirmed. If you want a second opinion on whether a tool or model your team is waiting on is actually ready — or just well-marketed — get in touch.