Two weeks before the implied June 30 deadline, Gemini 3.5 Pro is still in internal use and a limited enterprise preview. Google unveiled the model at I/O on May 19 with a promised June GA, a 2-million-token context window, and a reasoning mode called Deep Think positioned behind the $250-per-month Ultra subscription. Pro hasn't shipped. Flash has, and Flash is now carrying the entire 3.5 narrative.

That's a precarious place to be, because the frontier moved while Google was preparing slides.

Gemini 3.5 Flash went generally available as gemini-3.5-flash on June 1, per Google's API changelog, alongside Managed Agents in public preview and the antigravity-preview-05-2026 model string. Google's own numbers for Flash are genuinely strong: 76.2% on Terminal-Bench 2.1, 1656 Elo on GDPval-AA, 83.6% on MCP Atlas, 84.2% on CharXiv Reasoning. Google says Flash beats the previous-generation 3.1 Pro on coding and agentic work and acknowledges, in its own framing, that Flash regressed on the hardest reasoning. The bet is that Deep Think on 3.5 Pro closes that gap. The bet only pays if Pro ships in June.

The competitive backdrop is the part Google can't manage. On June 1, the same day Flash hit GA, Anthropic confidentially filed an S-1 with the SEC. The figures around that filing reset the conversation: a $965 billion valuation, ahead of OpenAI's $852 billion mark from late March, and annualized revenue that went from roughly $9 billion at the end of 2025 to $47 billion by May. The SpaceX prospectus disclosed Anthropic is paying $1.25 billion per month for compute through May 2029, much of it routed to Colossus 1 in Memphis.

Daniela Amodei, asked in a TechCrunch interview ahead of the IPO about doubts over AI's returns, shrugged them off. The revenue curve is the rebuttal. Claude Opus, meanwhile, still holds the SWE-bench lead Google would need 3.5 Pro to take.

Pricing for Pro is unconfirmed but expected to land near $15 per million input tokens and $60 per million output, tracking the prior generation's roughly 10x-Flash ratio. At that level, Pro has to clearly outperform Opus on agentic coding to justify the spread. Deep Think and a 2-million-token window are the differentiators on paper. Neither matters until external evaluators get hands on the model.

There's a version of this story where Google ships Pro on June 29, the benchmarks land, and the May 19 announcement reads in hindsight as confident sequencing. There's another where the window closes, Anthropic prices its IPO into the vacuum, and the 3.5 generation gets remembered as the one where Flash did the heavy lifting because Pro arrived late to a market that had already repriced.

Sources

https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-5/
https://ai.google.dev/gemini-api/docs/changelog
https://www.cnbc.com/2026/06/01/anthropic-ipo-s1-prospectus.html
https://techcrunch.com/2026/06/04/ahead-of-its-ipo-anthropics-daniela-amodei-shrugs-off-doubts-about-ais-returns/
https://www.techtimes.com/articles/317919/20260606/google-gemini-35-pro-nears-june-launch-2-million-token-context-deep-think-reasoning.htm