Update, August 7, 2026. This forecast missed. On June 18 I put Gemini 3.5 Pro's public release at July 1, with an 80% interval that ran out on August 6. That window has now closed and the model is still not out. I said at the end of the original piece that I would grade these against the release, so here is the grade, the reason I was wrong, and a new number. The capability and pricing forecasts further down have not resolved, and they stand as written.
The bar I was forecasting was the full public release, anyone in the Gemini app and through the API on standard Vertex and AI Studio plans. Gemini 3.5 Pro clears none of it today. It is missing from the public Gemini API model catalog, the pricing page, and Google Cloud's generally available and preview lists, and DeepMind's own model page still marks it coming soon. On July 21 Google shipped three new Gemini models and left the flagship out. A date forecast whose 80% interval expires with nothing shipped is simply wrong, and there is no partial credit for being wrong in the same direction as everyone else.
Why I got it wrong
I had the right worry and priced it far too low. The original piece said I could not rule out a much longer delay, but it pinned that risk on a safety hold driven by the Claude Fable drama. The real cause was more ordinary. Bloomberg reported on July 16 that the model was months behind schedule because its coding performance fell short of Google's own internal bar, and that a late-June refresh of the training data made results worse instead of better. Layers of internal stakeholders and competing teams each building their own coding tools slowed it further.
The deeper mistake was anchoring on a lag instead of on a cause. My reasoning was that DeepMind's on-stage teasers have historically trailed the full public release by a couple of weeks, so I took Pichai's "next month" and added a few. That works when the only thing between announcement and launch is release engineering. It is the wrong model when the model itself is not good enough yet, which is what was actually true in June. I even spotted the tell, that DeepMind shipped Flash at I/O while holding Pro back, and read it as caution about deployment. It was a capability problem.
I also ignored a base rate I had written down myself. I noted in the same piece that Google is notorious for announcing at I/O and shipping much later, sometimes never, and then drew an interval that allowed only six weeks past the promised month. Believing that base rate requires a far fatter right tail than the one I gave it.
Where the release date sits now
I re-ran the release question today with no steer toward any answer. The median is September 20, 2026, and the 80% interval runs from August 16 to December 15.
That is later than the prediction markets, which have clustered around a release this month. Two things explain most of the gap. Those contracts price a looser bar than mine, in some cases just the next Gemini Pro model rather than 3.5 Pro specifically, and they have already been wrong about the June and July deadlines on this same launch. Leaks point at Google's Made by Google event on August 12, and an announcement there is plausible. An announcement is not general availability in both the app and the standard-tier API, and DeepMind has staged rollouts behind previews and paid tiers before. A launch event followed by weeks of gated access would settle most market contracts and not this one.
Deep Think has not missed yet. Its June interval runs to September 7, so that call is still live, but the re-run pushes it well past the base model, to a median of October 25 with a right tail into April 2027. Part of that tail is a naming risk rather than a delay. Deep Think currently sits at version 3.1 while 3.6 Flash is already shipping, so Google could bridge to a Gemini 4 or 3.6 Pro release and never put out anything called 3.5 Deep Think at all.
Two other things have moved since June. On the July earnings call Pichai stopped giving dates for 3.5 Pro and talked instead about Gemini 4 pre-training and a near-monthly release cadence. Then on August 5 Google reshuffled its AI leadership, moving Demis Hassabis to chairman and chief scientist and elevating Koray Kavukcuoglu, with several senior researchers departing. Neither is a shipping signal. A reorganization that size usually adds friction to a launch that is already late, and the change in talking points fits reports that 3.5 Pro has been deprioritized in favor of Gemini 4. That risk sits in the right tail of the new forecast, not in its median.
The original June 18 analysis follows, unchanged except for the two release dates in the table.
DeepMind seems to me to be six to nine months behind the AI frontier, at least with respect to LLMs. At Google I/O on May 19, Sundar Pichai opened on what he called the agentic Gemini era, shipped the smaller Gemini 3.5 Flash, and said the flagship, Gemini 3.5 Pro, would arrive next month. On June 18 it still has not. I forecast it reaches the public in early July, a step behind the top of the field.
The reasoning and agentic coding Pichai showed on stage is roughly what Anthropic and OpenAI gave paying users earlier this year. DeepMind's best shipping model, Gemini 3.1 Pro, came out in February and trails today's leaders, and Deep Think, its strongest result, sits behind an Ultra subscription that runs up to $250 a month. Gemini 3.5 Pro is the model meant to close that gap, and it is late.
How good will it be?
I forecast Gemini 3.5 Pro at 160 on Epoch's Capabilities Index, an aggregate of about forty benchmarks, a point behind Claude Fable 5 at the top. Personally, I think Fable is significantly better than this eval shows, but this is still a useful target for forecasting. And anyway the strongest model anyone can actually run sits at GPT-5.5 Pro and Opus 4.8, about where I put 3.5 Pro, or maybe slightly ahead of it. This does depend on Deep Think, DeepMind's inference-time reasoning mode, which already posts 84.6% on ARC-AGI-2 and a gold medal at the 2025 IMO. Switch Deep Think off and the model drops a tier.
I put it behind the frontier rather than on it because of some FutureSearch forecasts of where it will land on specific benchmarks, shown below:
| Gemini 3.5 Pro | FutureSearch forecast |
|---|---|
| Public release, app and API | Sept 20, 2026 (Aug 16 to Dec 15), re-run Aug 7 |
| Deep Think mode, public | Oct 25, 2026 (Sept 5 to Apr 1, 2027), re-run Aug 7 |
| Epoch Capabilities Index, with Deep Think | 160 (156 to 162) |
| Artificial Analysis Intelligence Index | 61 (55 to 67) |
| Humanity's Last Exam, no tools | 49.9% (46 to 55) |
| Terminal-Bench, agentic coding | 83.8% (77 to 88) |
| Input price per million tokens | $4.33 ($2 to $13.50) |
| Output price per million tokens | $24.33 ($12 to $60) |
| Ships a 2 million token context window | 44% (1 million more likely, at 52%) |
| Keeps four named effort levels | 73% |
| No separate Ultra model, Deep Think stays a Pro mode | 85% |
| Share of OpenRouter coding traffic, first week | 2.5% (0.6 to 7) |
Ranges are 80% intervals. Percentages are the probability of the stated outcome. The two release dates were re-run on August 7 and replace the June numbers, which were July 1 (June 23 to August 6) and July 11 (June 24 to September 7). Everything else is as published on June 18.
What will Gemini-3.5-Pro be, and when will it ship?
Pichai said next month. I have it a few weeks later, in early July. The tell is what DeepMind did at I/O, shipping Flash while holding Pro back for more testing. (Google is notorious for vaporware announcements at I/O, frequently not shipping until the next calendar year, or sometimes not at all. Disclaimer: I worked at Google from 2014-2022.) The bar I am forecasting is the full public release, anyone in the app and through the API on standard Vertex and AI Studio plans, and that has trailed the on-stage teaser by a couple of weeks on past launches. Deep Think arrives later still, because DeepMind has kept its top reasoning mode in extra safety review before letting it out.
Forecast release window for Gemini 3.5 Pro against Google's I/O promise of next month.
What will it cost? Google models try to lead on price, and I expect it to bend that without breaking it. Gemini 3.1 Pro runs $2 and $12 per million tokens, and even a doubled 3.5 Pro lands at half the price of GPT-5.5 and Claude Opus, which keeps DeepMind the cheap seat at the frontier. The one path that breaks it is the $15 and $60 rumor going around, which would put DeepMind at parity and mark a real change of strategy. I give it weight in the tail, not the median.
The 2-million-token context window DeepMind has teased is closer to a coin flip than a promise, because it said 2 million for Gemini 2.5 Pro and shipped one. Two product calls look settled. The thinking control stays at four levels, since DeepMind's own developer docs already rule out the extra-high tier its rivals added. And there is no larger Ultra model coming. Ultra has become a subscription, and peak capability is Deep Think running on Pro.
A frontier-class model does not hand DeepMind back the developers it lost. Its share of usage on OpenRouter runs at about half of Anthropic's, and the coding category there belongs to Claude. I expect 3.5 Pro to take a sliver of that coding traffic in its first week, because the metric rewards cheap high-volume models and because most of DeepMind's coding usage never touches OpenRouter, running through its own tools instead. Underneath both, the people building agents have standardized on Claude Code. Winning them back takes a better model and time, and 3.5 Pro supplies only the first.
Where I might be wrong
It's hard to forecast capabilities, even though they seem continuous, it is possible Gemini-3.5-Pro is a Fable-class model. If Deep Think clears the frontier on math and science the way it cleared ARC-AGI, I may have undercalled it. I also can't rule out that it doesn't ship for a lot longer than I (or Sundar) think, because all the Fable drama could lead to a safety hold that pushes the release into August. Price is third. If the $15 and $60 numbers hold, DeepMind is fighting on capability instead of cost, a different company than the one these forecasts describe.
If 3.5 Pro ships in the next few weeks and the benchmarks land where I expect, DeepMind is back in the race, close behind the frontier. I argued in January that Anthropic was my pick for the top lab of 2026, and a 3.5 Pro this close to the frontier makes that race closer without settling it. For the financial side of the same three companies, see our OpenAI and Anthropic forecasts. I will grade these against the release.
About this forecast
Published June 18, 2026 with a median release date of July 1, 2026 and an 80% interval of June 23 to August 6, and a Deep Think median of July 11 with an interval of June 24 to September 7. Updated August 7, 2026, after the release interval expired with no launch. The release forecast is graded above as a miss and re-run, moving the median to September 20, 2026 with an 80% interval of August 16 to December 15. The Deep Think interval had not yet expired and was re-run rather than graded, moving to a median of October 25, 2026 with an 80% interval of September 5, 2026 to April 1, 2027. Both re-runs were made with neutral context, and the release question was run a second time with a deliberately pessimistic framing as a check; that version landed eight days earlier, at September 12, which is why the published number is the neutral one. The capability, price, and product forecasts are untouched since June 18 and resolve when the model ships.
Run this forecast yourself in the FutureSearch app and ask it to refresh the numbers the moment DeepMind ships.