Google's most anticipated AI model, Gemini 3.5 Pro, will not arrive in June as originally planned. The launch has slipped to July 2026, with Google citing continued refinement work — particularly around the model's complex reasoning performance. That pushes its standout features — a 2-million-token context window and an exclusive reasoning mode called Deep Think — just out of reach for another month.
What Gemini 3.5 Pro Was Supposed to Deliver
Announced at Google I/O on May 19, 2026, Gemini 3.5 Pro was positioned as Google's strongest deployed model to date. Two features defined it. First, a 2-million-token context window — double that of Gemini 3.5 Flash, and the largest of any production frontier model when announced. Second, Deep Think: a multi-step reasoning mode designed for problems that require sustained, structured logic rather than fast pattern matching.
Deep Think is not available across all tiers. It's gated behind Google's $250-per-month Ultra subscription — a pricing decision that generated significant discussion, as it concentrates the model's most capable features inside an enterprise paywall.
What's Already Shipping: Gemini 3.5 Flash
While Pro waits, Gemini 3.5 Flash has been generally available since May 19, and by most accounts it's a genuinely strong model for its tier. Flash is free in the Gemini app and AI Mode, and paid via the API. It outperforms Gemini 3.1 Pro on several coding and agentic benchmarks — Terminal-Bench 2.1 (76.2%), GDPval-AA (1656 Elo), and MCP Atlas (83.6%) — despite being the faster, cheaper model in the family.
A Pattern Across the Entire Industry
The delay is frustrating but not surprising. AI labs have increasingly settled into a rhythm of capability announcements that precede deployment by weeks or months: safety evaluations, infrastructure scaling, fine-tuning for deployment, and regulatory review all introduce friction that preview timelines don't account for.
OpenAI's GPT-5.6 is in a similar position, with prediction market odds for a June launch collapsing from ~83% to ~18% in the space of a week. Anthropic's Claude Fable 5 and Mythos 5 models launched in early June but were immediately restricted by a US government export-control directive. The public narrative of 2026 has been relentless AI acceleration; the operational reality is messier.
What to Watch in July
When Gemini 3.5 Pro does ship, the key question isn't the benchmark numbers — it's whether Deep Think's reasoning quality actually justifies the $250/month Ultra tier. Google has released no external evals for that mode, and at that price point, enterprise buyers will expect concrete evidence of performance before committing. The 2-million-token context window has clearer enterprise utility: processing entire codebases, large legal document sets, or research corpora in a single pass is valuable in ways that are straightforward to validate.
July is the new target. A second slip would be a real credibility issue for a company that positioned this as its flagship model moment of 2026.