Is Gemini 3.5 Pro still available in June 2026?

Not in general availability. As of June 25, 2026, Gemini 3.5 Pro remains in limited enterprise preview only, with general availability now targeted for July 2026 — past the original June 2026 commitment Google made at Google I/O on May 19, 2026. Limited-preview customers (enterprise testing partners and select Google Cloud accounts) can use it via Vertex AI under access controls, but the public consumer Gemini app and the standard Gemini API still default to Gemini 3.5 Flash (which did ship at I/O) for most queries. Pricing, model card, and broad rollout are expected to follow GA.

Why did Google delay Gemini 3.5 Pro?

Google's stated reason is 'quality refinements after early enterprise testing.' The unstated context is harder. (1) Two flagship AI researchers left Google DeepMind in one week: Noam Shazeer (co-author of the original Transformer architecture, Gemini co-lead) departed for OpenAI on June 18, and Nobel laureate John Jumper (AlphaFold co-creator) departed for Anthropic on June 19. (2) Demis Hassabis publicly responded on June 23 that DeepMind is 'still winning the talent war,' which is itself a signal that the perception had shifted. (3) Reports of internal quality concerns about Gemini 3.5 Pro's reasoning consistency on long-context tasks have surfaced from multiple enterprise testers. The delay is most likely a combination — real quality refinement plus difficult internal week.

What does Gemini 3.5 Pro do that Flash doesn't?

Gemini 3.5 Pro is the flagship of the 3.5 family — designed for complex reasoning, long-context understanding (the announced context window is 2M tokens), Deep Think reasoning mode, and agentic workloads. Gemini 3.5 Flash (the version that did ship at I/O) is the cost-optimized, latency-optimized sibling, intended for high-volume general queries, fast responses, and lower per-token cost. The pricing gap is meaningful — Pro is announced at roughly $15/$60 per 1M input/output tokens vs Flash at far lower. For most consumer chat and simple agent tasks, Flash is sufficient. Pro is for the complex-reasoning workloads where you would otherwise reach for Claude Opus 4.8, Claude Fable 5, or GPT-5.5 Pro.

Should I plan around Gemini 3.5 Pro for July 2026?

Be cautious. Google's June commitment slipped. The July window is plausible but not guaranteed. Three planning recommendations: (1) Don't build production workflows that critically depend on Gemini 3.5 Pro being live in July — keep Claude Fable 5, GPT-5.5 / GPT-5.6, and Gemini 3.5 Flash as fallback options. (2) If you have Google Cloud credits or enterprise commitments, request limited-preview access now via your Google Cloud account team — preview access is real even when GA slips. (3) Watch for the model card and pricing announcement — those typically appear 1-2 weeks before broad GA. As of June 25, neither has been published.

Quick Answer

Gemini 3.5 Pro Delayed to July 2026: What Happened

Published: June 25, 2026

Gemini 3.5 Pro Delayed to July 2026: What Happened

Google has pushed Gemini 3.5 Pro general availability from June 2026 to July 2026, citing quality refinements after early enterprise testing. The delay lands in a week where Google DeepMind also lost two flagship researchers (Noam Shazeer to OpenAI on June 18, John Jumper to Anthropic on June 19) and where Demis Hassabis publicly responded that DeepMind is “still winning” the AI talent war. Here’s what’s happening, what changes, and how to plan around it.

Last verified: June 25, 2026.

TL;DR

Original commitment: Gemini 3.5 Pro GA in June 2026 (announced at Google I/O on May 19, 2026)
Current status: Slipped to July 2026; only Gemini 3.5 Flash actually shipped in June
Google’s stated reason: “Quality refinements after early enterprise testing”
Wider context: Same week as Noam Shazeer (June 18) and John Jumper (June 19) departures from DeepMind
What works today: Gemini 3.5 Flash (GA), Gemini 3.5 Pro in limited enterprise preview only
Confirmed specs: 2M token context window, Deep Think reasoning, ~$15/$60 per 1M input/output tokens (announced, subject to GA confirmation)
What to do: Keep Claude Fable 5 / GPT-5.5 / Flash as your near-term default; plan Pro for late July at earliest

The timeline

Date	Event
May 19, 2026	Google I/O — Gemini 3.5 family announced; only Flash ships that day, Pro held in limited preview, GA promised for June 2026
June 2026 (mid-month)	Limited enterprise preview broadens slightly; quality-refinement reports begin from testers
June 18, 2026	Noam Shazeer (Gemini co-lead, original Transformer co-author) announces departure for OpenAI
June 19, 2026	Nobel laureate John Jumper (AlphaFold co-creator) announces departure for Anthropic
June 22, 2026	Reports surface that Gemini 3.5 Pro GA will slip past the June window
June 23, 2026	Demis Hassabis (Semafor interview) says DeepMind is “still winning AI talent” — implicit acknowledgment of the perception shift
June 25, 2026	GA confirmed pushed to July 2026 per multiple analyst trackers

Both June and July targets came from Google’s own framing — there is no firm public ship date on the new July window. “Late June and early July” appears in analyst tracking, but Google has not committed to a specific date.

Why this matters

Competitive timing

Gemini 3.5 Pro is meant to be Google’s answer to Claude Fable 5 (which shipped June 9, 2026) and the rumored GPT-5.6 (launch window June 22-28, 2026 per Polymarket). All three are positioned as flagship reasoning models with very long context windows.

A one-month slip in this market is meaningful. Claude Fable 5 has already shipped its credit-window pricing, exported its model card, and been benchmarked. GPT-5.6 is in launch window. If Gemini 3.5 Pro slips into late July, it lands in a market that has already moved.

The talent narrative

The Shazeer and Jumper departures in the same week as the delay are unrelated mechanically (the model release schedule is decided months ahead) but inseparable narratively. Both researchers were senior, public, and connected to Google’s AI identity (Transformer + AlphaFold).

Hassabis’s “still winning” response on June 23 acknowledges that the perception has shifted. Google can argue with the facts (DeepMind has roughly 4,000 researchers; two departures don’t change the org meaningfully) but the perception affects recruiting, internal morale, and external partner confidence.

If Gemini 3.5 Pro had shipped on time and benchmarked well, the talent story would be a footnote. With the delay, it becomes a paragraph.

Enterprise planning

Enterprises that committed to “Gemini 3.5 Pro by June” as a workload anchor now need to re-plan. Two practical questions:

Bridge: what runs the workload between now and Pro GA? Gemini 3.5 Flash is good enough for most use cases. Pro is for complex reasoning, very long context (2M tokens), and Deep Think workflows.
Confidence: is the July GA firm? Google’s June commitment slipped. The default assumption should be “July or later, with a model card 1-2 weeks before GA.”

What Pro actually offers (announced, pending GA)

From Google’s I/O announcement and limited-preview disclosures, the confirmed Gemini 3.5 Pro spec sheet:

Context window: 2M tokens (up from 1M for Gemini 2.5 Pro)
Reasoning mode: Deep Think — extended-thinking mode for complex multi-step reasoning, similar to Claude’s extended thinking and OpenAI’s o1/o3 reasoning
Multi-modal: native image + audio + video + text input, text + image output
Pricing (announced, not yet GA): ~$15 per 1M input tokens, ~$60 per 1M output tokens
Availability at GA: Vertex AI (Google Cloud), Gemini API, Gemini app (consumer)
Comparable models: Claude Fable 5 (1M context, $15/$75 per 1M), GPT-5.5 Pro, GPT-5.6 (1.5M context rumored)

The 2M context window is the headline number — it matches Gemini’s lineage advantage on long-context tasks. The Deep Think feature is the strategic answer to extended-thinking reasoning in the rival flagships.

Comparison: Gemini 3.5 Pro (delayed) vs alternatives

If you can’t wait for July GA, here’s the practical comparison for flagship reasoning workloads as of June 25, 2026:

Model	Status	Context	Pricing per 1M (in/out)	Best for
Gemini 3.5 Pro	Limited preview, GA July 2026	2M tokens	$15 / $60 (announced)	Long-context reasoning, Deep Think
Gemini 3.5 Flash	GA now	~1M tokens	Much cheaper	High-volume, general
Claude Fable 5	GA since June 9, 2026	1M tokens	$15 / $75	Code, agent loops, long reasoning
Claude Mythos 5	GA since June 9, 2026	Larger	Higher	Hardest reasoning
GPT-5.5 / GPT-5.5 Pro	GA now	400K+	Varies	General flagship, broad ecosystem
GPT-5.6	Launch window June 22-28, 2026	1.5M tokens (rumored)	Not announced	Agentic + browser testing

The practical default while you wait: Claude Fable 5 for code and reasoning, Gemini 3.5 Flash for cheap volume, GPT-5.5 for general use.

What to do today

If you’re an enterprise that planned around June 2026 Gemini 3.5 Pro

Request limited-preview access via your Google Cloud account team — preview is still real even with GA slipped.
Build a fallback to Claude Fable 5 or GPT-5.5 Pro for the next 6-8 weeks.
Don’t commit production migrations to Gemini 3.5 Pro until the model card publishes (typically 1-2 weeks pre-GA).

If you’re a developer choosing a flagship model right now

For long-context (>500K) reasoning, Claude Fable 5 is shipping and benchmarked.
For lowest cost at high volume, Gemini 3.5 Flash is shipping.
For broad ecosystem and tooling, GPT-5.5 / GPT-5.5 Pro is the safer default.

If you’re tracking the AI talent narrative

Watch Demis Hassabis’s public statements over the next 2-3 weeks — the pattern around “we are still winning” language tells you whether the recruiting story stabilized or worsened.
Watch Q2 2026 earnings (Alphabet reports late July) — the model-release schedule conversation will be a focus.
Watch for the Gemini 3.5 Pro model card — that’s the most concrete pre-GA signal.

Bottom line

Gemini 3.5 Pro is delayed from June to July 2026 GA, with no firm public ship date on the new window. Google’s stated reason is quality refinements; the unstated context is a difficult week of talent departures from DeepMind. The delay is meaningful but not catastrophic — Gemini 3.5 Flash is shipping, the announced 2M-token Pro spec is competitive, and the talent picture (despite the headlines) is much wider than two researchers.

If you can wait for late July, the model should be worth waiting for. If you can’t, the alternative options have never been better.