Google’s Gemini 3.5 Pro delay stretches past 3 months

Three months after CEO Sundar Pichai told a packed Google I/O audience to "give us until next month" for Gemini 3.5 Pro, the model remains unavailable — leaving Alphabet ↘ 0.36% without the top-tier AI system it promised developers and enterprise customers.

A Trail of Missed Deadlines

At Google I/O on May 19, Pichai said Gemini 3.5 Pro was "already being used internally" and would arrive in June. That deadline passed without a release, with a Polymarket contract settling at 97% "no release". Google DeepMind then targeted July 17 after scrapping the model's original base layer in favor of a deeper pre-training cycle. That date also came and went. On July 21, Google shipped three new Gemini models but not 3.5 Pro, with DeepMind product lead Logan Kilpatrick saying only that the team "hopes to land soon".hackernoon+4

An early August target, which Polymarket briefly priced at 82% likely, also failed to materialize. Forbes reported on August 13 that persistent coding and reliability problems, senior researcher departures, and a possible complete retraining had contributed to what the outlet called "the longest-awaited model of 2026". The Los Angeles Times, citing people familiar with the matter, reported that Google has been struggling to meet internal performance goals, particularly in coding.forbes+2

Prediction Markets See No Quick Resolution

As of late August, Polymarket traders give just a 10% chance of a release by month's end and do not clear 50% probability until September 30, with odds reaching 71% only by October 31. The contract pricing implies a 69% probability that no new Gemini Pro model ships before the end of September. More than $1.1 million has been wagered on the market.proactiveinvestors+1

What Google Has Left

The delay means Google's strongest publicly available model remains the Flash tier — currently Gemini 3.7 Flash, released in August — a lightweight model designed for speed and agentic tasks rather than the complex reasoning and coding benchmarks where Pro-class models compete. The attached PCMag comparison notes that Google's current Pro offering cannot match OpenAI's ChatGPT on coding and complex reasoning tasks. Without 3.5 Pro, Google lacks a direct answer to OpenAI's GPT-5.6 Sol, which PCMag describes as "vastly more intelligent" than Gemini's models for coding.pcmag+1

Google has not announced a model identifier, pricing, or official launch date for Gemini 3.5 Pro. The Gemini API model list still shows no `gemini-3.5-pro` entry, and the company's API pricing page carries no row for the model.eesel