Artificial Intelligence

Google's Flagship Model Is Late With No New Date

Google promised Gemini 3.5 Pro in June. It is months behind on coding, no new date exists, and Alphabet lost 200 billion dollars in a day. What that means for your roadmap.

Artificial IntelligenceBy Servola Tech Desk2026-07-174 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

Google's Flagship Model Is Late With No New Date

Key takeaways

Google said in May that Gemini 3.5 Pro would ship in June, the model is now months behind schedule because its coding results disappointed, and no replacement date has been given, so any plan of yours that names this model is resting on a date that does not exist.
Alphabet closed 4.4 percent lower on 16 July and shed roughly 200 billion dollars of market value on the report alone, which tells you the market now prices frontier delivery rather than frontier promises.
Google is shipping the cheap tier while the frontier tier slips: Flash is out and upgraded, Pro is not, so read a vendor's release pattern as the honest version of its roadmap.
If a named unreleased model sits on your critical path, replace the name with a capability and a tested fallback this week, and never sign a customer deliverable against a launch that has not happened.

The date came and went without a model

Google announced Gemini 3.5 Pro in mid-May at its developer conference and said the model would arrive the following month. June passed. Bloomberg reported on 16 July that the model is months behind schedule, and that Google is still taking time to improve it.

The sticking point is code. Google updated the data used to train Gemini late in June specifically to lift its coding skills, and, in the words of people familiar with the work, the results were disappointing. A Google spokesperson said the company is currently testing 3.5 Pro and an upgraded Flash model with partners.

No new launch date has been given. That is the fact worth carrying out of this story, because a missing date is not a small version of a date.

Two hundred billion dollars in one session

Alphabet closed down 4.4 percent on Thursday 16 July, ending the session at an unofficial 354.46 dollars and erasing roughly 200 billion dollars of market value. The decline dragged the wider indices to their session lows: the S and P 500 fell 0.51 percent to 7,533.77 and the Nasdaq Composite lost 1.47 percent to 25,881.95.

Nothing shipped that day and nothing broke. A report that a model is late moved a company by the size of a mid-cap index. The market has stopped paying for announced capability and started paying for delivered capability, and that repricing is the part of this story that outlives the model.

The release pattern is the real roadmap

Look at what Google has actually shipped in the 3.5 line: Flash, and an upgraded Flash now in testing. Gemini 3.5 Flash remains the only member of the series that customers can use, while the flagship stays in the lab. Gemini 3.1 Pro, still the current Pro, dates from February.

Google's own statement is the clearest signal here. Asked about a delayed frontier model, the company answered that it ships quickly across a wide range of models while keeping them highly cost-effective. That is a true sentence about the cheap tier, offered in place of an answer about the expensive one.

A vendor's release pattern is harder to spin than its roadmap, because a shipped model either exists or does not. Google is currently a vendor that reliably ships fast, cheap models and unreliably ships frontier ones. Plan against the pattern.

Ten employees, two rivals, one quarter

Ten current and former Google employees told Bloomberg they were concerned the company is losing ground while Anthropic and OpenAI ship models that outperform Gemini. That is not a market rumour, it is the people building the thing describing the gap from the inside.

The competitive frame matters less to an owner than the timing frame. Rivals released models that beat Gemini on code inside the same quarter in which Gemini's own coding work disappointed. Whatever the eventual quality of 3.5 Pro, the window in which it would have arrived first has already closed, and Google is now shipping into a comparison rather than setting one.

What to do before your next planning cycle

Strike unreleased model names from your plans and put capabilities in their place. A requirement that reads "a model that passes our code-review suite at 90 percent" survives a delay; a requirement that reads "Gemini 3.5 Pro" becomes a hole in a spreadsheet the moment a vendor slips.

Then price the delay you are already exposed to. Name the fallback you have actually tested on your own tasks, date the review, and tell any customer whose deliverable depends on an unreleased model before they read about it. The cost of a slipped launch is small when your plan expected it and large when your plan promised it.

The bottom line: Google will very likely ship a strong 3.5 Pro, and it will very likely be late again. Build the schedule that is true in both cases.

Frequently asked questions

When will Gemini 3.5 Pro actually launch?

Nobody outside Google knows, and Google has not said. It was announced in mid-May for a June launch, that date passed, and Bloomberg reported on 16 July that it is months behind schedule with no new date announced. Google says it is testing the model with partners, which is a status and not a date.

Is Gemini now behind OpenAI and Anthropic?

On coding specifically, the reporting says yes for now. Google's own attempt to lift Gemini's coding scores by retraining on updated data produced disappointing results, and ten current and former employees told Bloomberg they fear the company is losing ground while rivals ship models that outperform it. That is a statement about this quarter, not a permanent ranking.

Does the 4.4 percent drop mean Alphabet is in trouble?

No. It means the market reprices AI leadership on delivery rather than on announcements, and one report was enough to remove about 200 billion dollars of value in a session. Treat it as a signal about how fragile frontier expectations are, not as a verdict on the company.

A launch date is a forecast a vendor makes about itself, and this quarter has now broken several of them across several labs. Write your plans so that a model arriving late costs you a review meeting rather than a customer.

GoogleGoogle DeepMindGemini 3.5 ProAlphabetAI roadmapModel releasesAI procurement

More Than 15 Partners, and No Published Rules

Google DeepMind says it advanced more than 15 biosecurity partnerships in 12 months and named none of them. There are no published criteria, no application process and no appeal path. That is a procurement fact, not an ethics complaint.

6 min read

Law & Regulation

Vet a paid partner, lose the EU hosting shield

The Court of Justice held on 16 July 2026 that Google's pre-contract review of a YouTube channel means Article 14 hosting immunity does not apply. What owners who pay content partners should do now.

7 min read

Digital Sovereignty

The EU Council Revived the Rule That Scans Private Messages

The EU Council revived the lapsed rule that lets providers scan private messages with AI. For owners, confidential communication is now a governance choice.

2 min read

← All articles