If you searched “Gemini 3.5 Pro release date,” here is the honest status: there still isn’t one. Google promised the model “next month” at I/O 2026, missed June, missed mid-July, missed early August, and shipped four Flash models in the gap. This page is the running status record — checked against Google’s own model list, pricing page and changelog on the date above — plus what to actually run while the Pro tier stays empty.
Is Gemini 3.5 Pro released yet?
No. As of 23 August 2026, every one of the following is verifiable from Google’s own documentation:
- No
gemini-3.5-proentry exists in the Gemini API model list. The newest Pro-tier model isgemini-3.1-pro-preview. - No 3.5 Pro entry exists in the Vertex AI model garden either — 3.1 Pro is still the top Pro listing.
- No price appears on the Gemini API pricing page. Only 3.5 Flash and 3.5 Flash-Lite carry the 3.5 label there.
- No mention in the Gemini API release notes. The most recent entry is 13 August 2026 — Gemini 3.7 Flash going GA.
- No model card, no system card, no launch post.
What Google has published is a status label. The official Google DeepMind models page lists Gemini 3.1 Pro as the current Pro model and carries a “3.5 Pro coming soon” marker. That matters: the model is delayed, not cancelled. There is a difference, and a lot of secondary coverage blurs it.
The awkward part for Google’s naming scheme is that the best Gemini model you can actually call today is not a Pro model at all. It’s Gemini 3.7 Flash, which went GA on 13 August.
Is “gemini-3.5-pro-preview” a real model you can call?
Not publicly — and this is worth spelling out, because the exact string gemini-3.5-pro-preview has become a common search in its own right.
The string follows Google’s standard convention: Pro models ship first under a -preview suffix (gemini-3.1-pro-preview, gemini-3-pro-preview) before a stable ID appears. So it is a reasonable guess at what the eventual ID will be. But as of 23 August 2026 it is a guess, not a live endpoint:
- It is absent from every published model list, on both AI Studio and Vertex AI.
- Developers have filed allowlist requests for it on Google’s own AI Developers Forum — one on 24 July 2026 — and Google’s reply redirected the requester to their sales rep and the Cloud forum without confirming or denying that the model exists.
- There is no public preview signup, no eligibility policy, and no documented endpoint.
What does exist is partner testing. Google’s on-record line since 16 July has been that it is “currently testing 3.5 Pro… with partners,” and Gemini API product lead Logan Kilpatrick repeated on 21 July that the model is testing with partners and would “hopefully land soon.” Several outlets describe this as a “limited Vertex AI enterprise preview.” That framing is reasonable, but it is not a tier you can request — treat any site offering you 3.5 Pro API access today as either mislabelling 3.1 Pro or making it up.
What happened with Gemini 3.5 Pro in August 2026?
August was the month the delay stopped being a product story and became an org story. Three dated events, in order:
5 August — Google DeepMind leadership shake-up. Alphabet announced that Demis Hassabis would step down as CEO of Google DeepMind, moving to chairman of DeepMind and chief scientist of Alphabet. CTO Koray Kavukcuoglu took over day-to-day operations as SVP, reporting directly to Sundar Pichai and based in Mountain View rather than London — which puts the Gemini roadmap, including 3.5 Pro, under new management. On the same day, Jeff Dean announced his departure after 27 years to co-found Discovery Loop, a public benefit corporation with Alphabet as a founding investor. Fortune reported Alphabet shares fell about 5% on the news, and named 3.5 Pro being “months behind its original June launch target” as part of the backdrop.
10 August — the causal link gets reported. A follow-up Fortune investigation tied the reshuffle directly to stalled models and missed deadlines. Three DeepMind engineers told Fortune the company had underprioritised coding. One put it: “They dropped the ball a little bit [on coding]… By the time you realize something like that, a number of months have passed, and now you have to play catch-up.”
13 August — Gemini 3.7 Flash ships; Pro doesn’t. Google’s changelog entry calls 3.7 Flash “our most intelligent workhorse model yet for coding and agents.” Pro went unmentioned. Forbes covered it as the delay continuing, describing 3.5 Pro as the “longest-awaited model of 2026” and citing a report that Google may have restarted foundational pre-training because the issue was “a structural problem, not a finishing problem.”
And 18 August? If you searched that specific date, here is what actually happened: the Gemini Enterprise release notes logged Gemini 3.6 Flash becoming available in US and EU multi-regions with the US allowlist requirement dropped, alongside a minor home-page feature. Nothing about 3.5 Pro. It was a Flash rollout note, not a Pro launch.
Why is Gemini 3.5 Pro delayed?
Coding performance, then reliability, then possibly the foundations. The original reporting is Bloomberg’s on 16 July: Google was “taking time to try to improve [Gemini 3.5 Pro’s] capabilities, particularly in coding,” after it updated the training data in late June to fix coding skills and “the results were disappointing.” Alphabet stock dipped on the report.
The full slip trail:
| Date (2026) | What happened to Gemini 3.5 Pro |
|---|---|
| 19 May | Announced at I/O: “already being used internally, and we look forward to rolling it out next month.” Gemini 3.5 Flash ships the same day. Pro doesn’t. Target #1: June. |
| Late June | June target missed. Training data updated to improve coding; results reported as disappointing. Slips to July. Target #2: 17 July. |
| 16 July | Bloomberg reports the delay is about coding performance. Google: “currently testing 3.5 Pro, an upgraded Flash model, and other models with partners.” No new date. |
| 21 July | Google ships 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash Cyber — no Pro. Logan Kilpatrick: testing with partners, “hopefully land soon,” and separately announces the “most ambitious pre-training run yet, for Gemini 4.” |
| Early Aug | Target #3 missed. |
| 5 Aug | Hassabis steps back; Kavukcuoglu takes over day-to-day and the Gemini roadmap; Jeff Dean departs. Alphabet down ~5%. |
| 13 Aug | Gemini 3.7 Flash GA. Pro unmentioned. Reports of a possible foundational retrain. |
| 23 Aug | Still “coming soon” on DeepMind’s models page. No ID, no price, no date. |
Read the language progression rather than the dates: “next month” → “July” → “testing with partners” → “as soon as it’s ready” → “hopefully land soon” → silence. Each statement is weaker and less falsifiable than the one before. Any specific 3.5 Pro release date you see quoted today — including the “12 August” one that circulated widely — is speculation that has already expired.
One more signal worth weighing: Google publicly started Gemini 4 pre-training on 21 July. A lab that has moved its most ambitious training run to the next generation is a lab whose current-generation flagship has a shrinking window to matter.
What was Gemini 3.5 Pro supposed to include?
Two headline features were announced at I/O and have never been walked back — but also never shipped, so treat them as intent, not spec:
- A 2M-token context window. Double the 1M on Gemini 3.5 Flash, and it would have been the largest on any production frontier model at announcement. Nothing in the Gemini family offers 2M today — 3.1 Pro and 3.7 Flash are both 1M. If you planned an architecture around a 2M window arriving in June 2026, re-plan around 1M plus retrieval.
- Deep Think reasoning mode. Extended inference-time compute for hard science, maths and coding problems. Notably, this one didn’t wait for 3.5 Pro: DeepMind’s models page already lists Gemini 3.1 Deep Think as a shipping model “for modern challenges across science, research and engineering.” If Deep Think was the reason you were waiting, you can stop waiting.
Everything else you may have seen — projected benchmark scores, “$2–3 input” pricing estimates — was extrapolation from the 3.1 generation published before the delay. With late-June training runs reportedly disappointing and a possible foundational retrain since, those extrapolations have no basis. We’ve retired them from this page rather than re-dating them.
What should you use instead of Gemini 3.5 Pro?
Two answers, depending on what you were waiting for:
- For most work: Gemini 3.7 Flash. Google’s strongest available model, with a 1M-token context and introductory pricing that undercuts the Pro line by a wide margin.
- For Pro-tier API requirements: Gemini 3.1 Pro (preview). Still the Pro flagship. Choose it when something in your stack specifically requires the Pro line rather than raw capability.
Current Google pricing, taken from the official pricing page on 23 August 2026:
| Model | Context | Input / Output per 1M tokens | Status |
|---|---|---|---|
| Gemini 3.5 Pro | 2M (announced, unshipped) | never announced | Unreleased — “coming soon” |
| Gemini 3.7 Flash | 1M | $0.75 / $3.75 through 31 Dec 2026, then $1.50 / $7.50 | GA (13 Aug) |
| Gemini 3.6 Flash | 1M | $0.75 / $3.75 through 31 Dec 2026, then $1.50 / $7.50 | GA (21 Jul) |
| Gemini 3.5 Flash | 1M | $1.50 / $9.00 | GA (19 May) |
| Gemini 3.5 Flash-Lite | 1M | $0.30 / $2.50 | GA (21 Jul) |
| Gemini 3.1 Pro (preview) | 1M | $2.00 / $12.00 (≤200K); $4.00 / $18.00 (>200K) | Current Pro flagship |
Note the pricing inversion this creates: 3.7 Flash costs $0.75 in / $3.75 out while 3.1 Pro costs $2.00 / $12.00. You are paying roughly 2.7× on input and 3.2× on output for the Pro badge on an older generation. Unless you have a concrete reason to be on the Pro line, that spread is hard to defend. For a worked cost-per-task comparison against a cheaper alternative, see our DeepSeek V4 vs Gemini 3.5 cost breakdown.
How does the rest of the frontier compare right now?
The competitive context is the real reason the delay stings. While Google’s Pro tier sat empty, rivals shipped and repriced:
| Model | Input / Output per 1M tokens | Note |
|---|---|---|
| Claude Sonnet 5 | $2.00 / $10.00 | A planned 1 Sept rise to $3/$15 was cancelled; $2/$10 became standard on 15 Aug |
| GPT-5.6 Sol | $5.00 / $30.00 | Top tier |
| GPT-5.6 Terra | $2.00 / $12.00 | Mid tier — same list price as Gemini 3.1 Pro |
| GPT-5.6 Luna | $0.20 / $1.20 | Cheap tier |
| Grok 4.5 | $2.00 / $6.00 | Cheapest output among the $2-input flagships |
| Gemini 3.1 Pro | $2.00 / $12.00 | Google’s only Pro option |
Gemini 3.5 Pro is the only announced flagship among the major labs with no public access tier at all. And on the reporting around the 21 July release, 3.6 Flash was seen as underperforming rivals like Grok 4.5 and GPT-5.6 Luna on several benchmarks — which is precisely the gap 3.5 Pro was meant to close. If you’re evaluating across labs, our Claude Opus 5 launch guide covers the benchmark side of the current frontier in detail.
How will you know when Gemini 3.5 Pro actually ships?
Ignore rumour aggregators and watch four surfaces that change only on a real launch. In rough order of speed:
- The Gemini API release notes — every GA and preview lands here with a date.
- The model list — a
gemini-3.5-proorgemini-3.5-pro-previewID appearing is the first hard confirmation. - The pricing page — no model is really usable until it’s priced.
- The DeepMind models page — the “3.5 Pro coming soon” label flipping to a real card.
A practical hedge in the meantime: keep the model ID in config rather than hard-coded. The whole Gemini family shares one SDK surface, so switching is a one-line change whenever Pro lands. That advice has aged well — teams that shipped on Flash in May have upgraded twice since; teams that blocked on Pro have been waiting three months.
Pillar guide
For the whole Gemini family — every shipped model, pricing, benchmarks, and how it compares to Claude and GPT — see our Gemini 3.5 complete guide for 2026.
When will Gemini 3.5 Pro launch?
Nobody outside Google knows, and the honest reading of the evidence is that Google may not know either. Three missed targets, a reported foundational retrain, a leadership change that moved the Gemini roadmap to a new owner in a new time zone, and a public pivot of the most ambitious pre-training run to Gemini 4 — none of that describes a model that is days away.
So the decision rule is simple: build as if it isn’t coming, and treat it as upside if it does. If your use case needs Google, 3.7 Flash is the strongest thing on the menu and the cheapest per token in the family’s upper half. If it needs a 2M context window, that product does not exist at any lab today — architect around retrieval instead of waiting. And if you were holding out specifically for Deep Think, Gemini 3.1 Deep Think already shipped.
This page is maintained as a status record and updated when Google’s own surfaces change. If you’re staffing a team that has to make these evaluate-and-migrate calls quickly, Codersera matches you with vetted remote developers who work on production AI systems day to day.
FAQ
Is Gemini 3.5 Pro released yet?
No. As of 23 August 2026 there is no gemini-3.5-pro model ID in the Gemini API or Vertex AI, no pricing, no model card and no launch post. Google DeepMind’s official models page still labels it “3.5 Pro coming soon.” It was announced at I/O on 19 May 2026 for a June release that never happened.
When is Gemini 3.5 Pro coming out?
There is no announced date. Google’s last public positions were “as soon as it’s ready” (21 July) and Logan Kilpatrick’s “hopefully land soon.” The model has missed three targets — June, mid-July and early August 2026 — and was not mentioned at the 13 August Gemini 3.7 Flash launch. Any specific date circulating now is speculation.
Is gemini-3.5-pro-preview a real model string?
Not as a public endpoint. It matches Google’s naming convention for preview Pro models, and developers have filed allowlist requests for it on Google’s AI Developers Forum, but it appears in no published model list on AI Studio or Vertex AI and Google has not confirmed it. Only partner testing exists today.
Why is Gemini 3.5 Pro delayed?
Coding performance, per Bloomberg’s 16 July reporting: Google updated the training data in late June to improve coding and the results were disappointing. Later coverage points to reliability problems and a possible restart of foundational pre-training. Fortune reported DeepMind engineers saying the company had underprioritised coding and had to play catch-up.
What happened with Gemini 3.5 Pro on 18 August 2026?
Nothing. The 18 August Gemini Enterprise release note covered Gemini 3.6 Flash expanding to US and EU multi-regions with the US allowlist requirement removed, plus a minor home-page feature. It was a Flash availability update, not a 3.5 Pro launch, and no Pro model was mentioned.
What should I use instead of Gemini 3.5 Pro?
Gemini 3.7 Flash for most workloads — it is Google’s strongest available model at $0.75 input / $3.75 output per 1M tokens through 31 December 2026, with a 1M-token context. If you specifically need the Pro line, Gemini 3.1 Pro (preview) remains the Pro flagship at $2.00 / $12.00 up to 200K tokens.
What is the Gemini 3.5 Pro context window?
A 2M-token window was announced at I/O 2026 but has never shipped or been confirmed on a spec sheet. No Gemini model offers 2M tokens today — Gemini 3.1 Pro and Gemini 3.7 Flash are both 1M. Plan around 1M plus retrieval rather than waiting for 2M.
How much will Gemini 3.5 Pro cost?
No price has ever been published, and it does not appear on Google’s pricing page. The nearest reference point is Gemini 3.1 Pro at $2.00 input / $12.00 output per 1M tokens up to 200K context, rising to $4.00 / $18.00 above it. Any specific 3.5 Pro price quoted elsewhere is invented.
Has Gemini 3.5 Pro been cancelled?
No. Google DeepMind’s models page still carries a “3.5 Pro coming soon” label, and Google has repeatedly said it is testing with partners. Delayed and cancelled are different states. That said, Google began pre-training Gemini 4 on 21 July 2026, so the window in which 3.5 Pro is competitively relevant is narrowing.
Did the DeepMind leadership change affect Gemini 3.5 Pro?
It moved ownership. On 5 August 2026 Demis Hassabis stepped back to chairman of Google DeepMind and Alphabet chief scientist, with CTO Koray Kavukcuoglu taking day-to-day control and the Gemini roadmap as SVP reporting to Sundar Pichai. Jeff Dean left the same day. Fortune tied the reshuffle directly to stalled models and missed deadlines.