The situation as of 7 July 2026: Gemini 3.5 Pro was announced at Google I/O on 19 May with a June general-availability target. It missed June 30. It entered July still in limited Vertex AI enterprise preview. As of today — the second full week of July — there is no confirmed GA date, no published benchmarks, no final pricing, and no public API access.

What Google has confirmed

Google has reaffirmed a July 2026 target for Gemini 3.5 Pro general availability. The model remains in limited Vertex AI enterprise preview, available to select partners under a non-disclosure agreement. Google has not committed to a specific date within July, and based on the pattern from the June deadline miss, a further slip into August cannot be ruled out.

The confirmed capabilities — which have not changed since I/O — are:

  • 2,000,000-token context window (double the nearest competitor)
  • Deep Think reasoning mode (Google's extended-thinking equivalent)
  • Native Google Workspace integration (Gmail, Docs, Sheets, Drive)
  • Vertex AI multimodal input support
  • Indicated pricing: approximately £2.50 per million input tokens, £15 per million output tokens

Why it keeps slipping

The June 30 miss was attributed to "quality refinements after enterprise testing feedback" — a pattern that suggests the model passed capability benchmarks but failed on reliability or consistency metrics at enterprise scale. The most common failure mode for large-context models in enterprise testing is context degradation: performance at 200k tokens does not necessarily hold at 1 million or 2 million tokens.

Separately, four of Google's most senior AI researchers left for Anthropic in a six-day period in late June. While individual talent moves do not directly delay product launches, the concentration of that departure — including a Nobel Prize laureate and a transformer co-inventor — adds context to why Google may be taking additional time before releasing a flagship model under heightened scrutiny.

What UK businesses should do while waiting

If your use case is large-document processing (long contracts, tender packs, multi-chapter specifications), the 2 million token context window is genuinely useful. But you do not need Gemini 3.5 Pro to start building the workflow.

Three practical moves:

  • Test Claude Sonnet 5 now. It became Anthropic's default model on July 1 and has a 1 million token context window in Claude Code. Run your actual large-document task on it today and time the result.
  • Enable Gemini 2.0 Flash in Workspace. If you use Google Workspace, the Gemini sidebar is already available in Gmail and Docs. It handles most common document tasks and will transition to Gemini 3.5 Pro automatically when it launches.
  • Write your test case. Define exactly which task you want to run on Gemini 3.5 Pro when it launches — which documents, what output you need, what success looks like. You will be able to evaluate the model in thirty minutes instead of three days.

Operator decision: July 7

If you were waiting for Gemini 3.5 Pro to start an AI project: Stop waiting. Run the project on Claude Sonnet 5 or Gemini Flash today. You will have a real baseline when Gemini Pro launches — and you will not have lost another two weeks.
If you are evaluating AI platforms: Do not extend vendor contracts waiting for Pro. Evaluate the models currently available. Pro can be assessed when it ships, and the test case you have built will make the evaluation fast.
If Gemini Pro is not relevant to your business: Claude Sonnet 5 is available now on every Anthropic plan. That is likely the most practical AI tool for UK service businesses regardless of what Google ships this month.