Last updated: July 7, 2026

Gemini 3.5 Pro is the unreleased Pro-tier model people are watching after Google launched Gemini 3.5 Flash at I/O 2026. The problem is that search results, YouTube thumbnails, Reddit posts, and scraped “model pages” are already talking as if 3.5 Pro has a fixed launch date, final context window, and published benchmarks. That is not what the official sources show yet.

Quick answer: Google has confirmed that Gemini 3.5 Pro is coming, and Google executives said it was being used internally and expected after Gemini 3.5 Flash. But as of this source check on July 7, 2026, the public Gemini API pages list gemini-3.5-flash and gemini-3.1-pro-preview, not a generally available gemini-3.5-pro model ID. Treat every claim about a July 17 launch date, 2M context window, benchmark score, pricing table, or “scrapped base model” as unconfirmed until Google publishes a model card, API docs, pricing, or a release note.

Source check Sources checked

Verified July 7, 2026

Rumor disclaimer: the rumor section below is intentionally labeled. It is useful for watchlists, not for procurement, production routing, pricing estimates, or benchmark claims.

Gemini 3.5 Pro status at a glance

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.
Topic Status on July 7, 2026 Confidence
Gemini 3.5 Flash availability Generally available across multiple Google surfaces and API routes High
Gemini 3.5 Pro existence Confirmed as “coming” / “coming soon” High
Public gemini-3.5-pro API model ID Not found in the official API model/deprecation pages checked High
General availability date Not officially announced High
Reported June-to-July delay Reported by Business Insider; Google declined comment in that report Medium
July 17 launch date Rumor from community/leak ecosystem Low
2M context window Rumor / repeated claim, not confirmed in official Pro docs Low
Final pricing Unknown High
Final benchmarks Unknown High

The safest headline is: Gemini 3.5 Pro is real as a roadmap item, but not yet a public model you should write production docs around.

What Google has officially confirmed

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

Google’s official Gemini 3.5 announcement focuses on Gemini 3.5 Flash. Google describes 3.5 Flash as a model designed for agentic workflows, coding, long-horizon tasks, and fast iteration. It is already the model with official API documentation and published capability tables.

The official Google/DeepMind pages confirm these points:

  • Gemini 3.5 is a model family, not only one model.
  • 3.5 Flash is available now in Google products and developer surfaces.
  • 3.5 Pro is coming, but public details are limited.
  • Google is positioning Gemini 3.5 around frontier intelligence with action: agents, coding, multimodal understanding, and enterprise workflows.
  • 3.5 Flash is the visible proof point for the family while Pro remains unreleased or restricted.

On the Google DeepMind Gemini model page, the important phrase is simple: “3.5 Pro coming soon.” That is enough to confirm that Pro is on the roadmap. It is not enough to confirm a release date, model ID, or pricing.

Gemini 3.5 Pro release date: what happened?

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

The release-date story has three layers.

Official layer

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

At I/O 2026, Google said it was working on Gemini 3.5 Pro and expected to roll it out after Flash. In official I/O coverage, Google said 3.5 Pro was being used internally and was showing improvements.

That means Google set the expectation that Pro would follow Flash. But official pages checked here do not give a final public launch date.

Reported layer

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

Business Insider reported that Google had originally planned Gemini 3.5 Pro for June, then shifted the target to July while gathering feedback from early testers and tuning the model. The report says the model had been available to some users through Antigravity and LMArena-style testing contexts.

That is reporting, not an official Google release note. It is stronger than a random screenshot, but it should still be described as reported.

Rumor layer

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

The July 17 date appears in community posts, YouTube discussions, and leak-style summaries. It might turn out to be close. It might also be wrong, or it might refer to a preview, limited access, LMArena appearance, Antigravity test, internal milestone, or staggered rollout rather than general availability.

So the practical answer is:

Gemini 3.5 Pro was expected after Flash, reportedly slipped from June to July, and has a rumored July 17 target — but Google has not published a final GA date in the official sources checked here.

Gemini 3.5 Flash vs Gemini 3.5 Pro

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

Until 3.5 Pro launches publicly, the comparison is partly inferred.

Area Gemini 3.5 Flash Gemini 3.5 Pro expectation
Status Available Coming soon / not broadly public
API model ID gemini-3.5-flash No public gemini-3.5-pro ID confirmed in checked docs
Positioning Fast frontier-level model for agents and coding Expected Pro-tier model for higher-quality reasoning and complex work
Pricing Published in Google pricing surfaces Unknown
Benchmarks Published for Flash Not final / not official yet
Best current use Agentic loops, coding iteration, multimodal tasks, enterprise workflows Wait for official docs before routing production workloads

The key mistake is comparing rumored Pro against real Flash as if both are available. A better workflow is to evaluate 3.5 Flash today, keep 3.1 Pro Preview as the current Pro-tier reference, then rerun tests when 3.5 Pro gets a real model page.

Rumor tracker: what people are claiming

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

This is the section to read with the strongest disclaimer. These claims are circulating, but they are not confirmed in the official Google docs checked here.

Rumor What people mean How to treat it
July 17 launch A mid-July target for public or preview release Plausible but unconfirmed. Do not schedule production launches around it.
2M context window Pro may double the 1M-token context seen in current Gemini models Unconfirmed. Wait for model card/API limits.
Base model scrapped or retrained Google may have changed the underlying training strategy before release Very speculative. Could be misunderstood internal/testing chatter.
Better math Pro may target hard math gaps versus Flash and rivals Reasonable expectation, not a confirmed benchmark result.
Better SVG/frontend coding Google may be tuning for visual code, UI generation, and web demos Fits Google’s demos, but exact Pro capability is unconfirmed.
Image-generation companion launch Pro could arrive near image/video model updates Pure roadmap speculation until Google posts launch details.
Extra-high thinking mode Pro may expose a deeper reasoning tier Unconfirmed naming and product behavior.
LMArena or Antigravity stealth tests Some testers may have touched a hidden or early model Possible, but not equivalent to GA access.

A rumor can be useful without being reliable. For planning, convert each rumor into a test question:

  • If context is 2M, does quality actually improve over RAG?
  • If Pro improves frontend code, does it pass your design-system constraints?
  • If it is stronger at math, does it solve your actual data/science tasks?
  • If it launches in July, will rate limits and availability support your workload?

The 2M context-window rumor

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

The loudest technical rumor is that Gemini 3.5 Pro may support a 2-million-token context window.

Why people care:

  • it would fit Google’s long-context reputation;
  • it would help with codebases, legal files, research papers, and enterprise archives;
  • it would be easy to market against models with smaller context windows;
  • it would pair well with agents that need to keep more state across a workflow.

Why you should be cautious:

  • long context does not automatically mean better retrieval;
  • huge prompts can be slower and more expensive;
  • attention quality can degrade across very large inputs;
  • good RAG can outperform naive “stuff everything into context” workflows;
  • final token limits, caching rules, and pricing matter as much as the headline number.

If Google confirms 2M context, the right test is not “Can it accept a giant prompt?” The right test is: Can it find the right fact, use it correctly, avoid distractors, and produce a better answer than a cheaper retrieval pipeline?

Why Gemini 3.5 Pro matters

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

Gemini 3.5 Pro matters because Google is trying to turn Gemini from a chat model into an agent platform.

The public 3.5 Flash launch already points in that direction:

  • agentic coding;
  • computer use;
  • long-horizon tasks;
  • multimodal understanding;
  • enterprise workflows;
  • Google Antigravity integration;
  • Search and app experiences powered by agents.

If Flash is the fast worker, Pro is expected to be the deeper planner: better reasoning, stronger complex-task reliability, and more careful orchestration. That is why the rumor cycle is so intense. Developers want to know whether Gemini 3.5 Pro will be a better default for:

  • coding agents;
  • UI and frontend generation;
  • long document analysis;
  • research synthesis;
  • enterprise automation;
  • multimodal reasoning;
  • math-heavy or science-heavy prompts.

But until Google publishes official specs, the answer is: test Flash now, prepare a benchmark suite, and wait for Pro docs before making the switch.

What should teams do before launch?

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

Do not wait passively. Build the evaluation harness now.

Prep step Why it matters
Collect 20-50 real tasks Viral prompts do not represent your workflow.
Include easy, medium, and hard cases A Pro model may only justify itself on the hard tail.
Track cost per successful task Higher model quality is irrelevant if cost explodes.
Measure latency and retry count Agentic workflows can fail through slowness, not only bad answers.
Include long-context distractors This tests whether context is useful, not just large.
Run against Gemini 3.5 Flash today Establish a baseline before Pro appears.
Re-run after official Pro release Avoid trusting leaked benchmark claims.
Evaluation template Gemini 3.5 Pro launch checklist
Check What to verify Why it matters
Model ID Is there an official gemini-3.5-pro or preview slug? Prevents fake endpoint docs.
Access AI Studio, Gemini API, Antigravity, Enterprise, Gemini app Different surfaces may roll out at different times.
Context Official input/output limits Long-context rumors are not enough.
Pricing Input, output, cached input, priority/batch pricing Token economics drive routing.
Benchmarks Official and independent evals Avoid launch-day leaderboard overreaction.
Safety/model card Limitations, intended use, restrictions Especially important for enterprise and regulated work.

Should you wait for Gemini 3.5 Pro?

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

Wait only if the task specifically needs what Pro is likely to improve.

Use Gemini 3.5 Flash now when:

  • you need fast agentic coding loops;
  • you want to test Gemini’s current 3.5 direction;
  • latency and cost matter;
  • you are building subagents or high-volume workflows;
  • you can tolerate model limitations and rerun tests later.

Wait for Gemini 3.5 Pro when:

  • your task is bottlenecked by reasoning quality;
  • failures are expensive;
  • you need stronger long-context reliability;
  • you are comparing against top Pro/flagship models;
  • the current Flash/3.1 Pro setup is not good enough.

Do not wait when:

  • the rumor is your only reason;
  • you need production availability today;
  • your bottleneck is workflow design, not model quality;
  • you have not tested the current Gemini models yet.

Bottom line

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

Gemini 3.5 Pro is one of the most important unreleased AI models to watch in July 2026, but the public facts are still limited.

What is confirmed:

  • Google launched Gemini 3.5 Flash.
  • Google says Gemini 3.5 Pro is coming.
  • Official API docs currently show Flash and 3.1 Pro Preview, not a public 3.5 Pro model ID.
  • A delay from June to July has been reported, but Google has not published a final GA date.

What is not confirmed:

  • July 17 public launch;
  • 2M context;
  • final pricing;
  • final benchmarks;
  • exact API model ID;
  • new thinking mode names;
  • whether rumors about retraining or a scrapped base model are accurate.

The smart move: treat Gemini 3.5 Pro as a watchlist item, not a migration target, until Google publishes the model card, API docs, pricing, and availability details.

FAQ

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.
Is Gemini 3.5 Pro released?
Not as a generally available public API model in the official sources checked on July 7, 2026. Google has released Gemini 3.5 Flash and lists Gemini 3.5 Pro as coming soon.
What is the Gemini 3.5 Pro release date?
Google has not published a final general-availability date in the sources checked here. Google said Pro was coming after Flash, and Business Insider reported a slip from June to July. A July 17 date is a rumor, not an official date.
Will Gemini 3.5 Pro have a 2M-token context window?
That is a popular rumor, but it is not confirmed in the official Pro docs checked here. Wait for Google’s model card or API model page before relying on any context-window number.
What is the Gemini 3.5 Pro API model ID?
No official public gemini-3.5-pro model ID was found in the checked API model/deprecation pages. The currently documented 3.5 model is gemini-3.5-flash.
How is Gemini 3.5 Pro different from Gemini 3.5 Flash?
Gemini 3.5 Flash is the available fast model for agents and coding. Gemini 3.5 Pro is expected to be the higher-capability Pro-tier model, but final specs, pricing, and benchmarks are not public yet.
Should I trust Gemini 3.5 Pro leaks?
Use leaks as a watchlist, not as facts. Treat them as questions to verify when Google publishes official docs: context size, pricing, reasoning modes, benchmarks, latency, and API availability.