Google’s next flagship model missed its own internal bar for coding, and the market priced in the delay immediately. Alphabet stock dropped 4% on Thursday after Bloomberg reported, citing people familiar with the matter, that Gemini 3.5 Pro will not ship on the timeline Google set in May. The stakes are straightforward: a frontier lab that slips its coding benchmark window during the industry’s most competitive coding cycle to date risks losing developer mindshare it cannot easily win back.
Google unveiled Gemini 3.5 Pro at its I/O developer conference in May, telling attendees the model was already running internal workloads and would reach broader availability the following month. That month came and went without a public launch. Bloomberg’s sourcing points to a specific bottleneck: the model’s code generation fell short of what Google’s own teams expected, prompting engineers to keep tuning rather than ship.
An Alphabet spokesperson told CNBC the company is “shipping quickly across a wide range of models while keeping them highly cost-effective for customers,” and confirmed that Gemini 3.5 Pro, a faster Flash update, and further launches all remain in partner testing. That statement confirms the delay without disputing Bloomberg’s reporting on the cause. Google did not offer a revised launch date.
The timing damages Google’s coding narrative specifically. Meta launched Muse Spark 1.1 last week, and AI chief Alexandr Wang called it the company’s strongest model yet for agentic and coding work. OpenAI shipped GPT-5.6 Sol in the same week; CEO Sam Altman said it runs 54% more token efficient on agentic coding tasks than its predecessor. Both moves reset the coding leaderboard days before Google’s own upgrade stalled, and code generation has become the primary battleground where Anthropic, OpenAI, and now Meta compete for developer accounts and enterprise contracts.
Coding performance now functions as the industry’s shorthand for whether a model is good enough to matter, because agentic coding tools are where enterprises spend budget first. A lab that cannot clear that bar loses pull with the developers who choose which API becomes the default inside a company’s tooling. Google has not disclosed a benchmark score for Gemini 3.5 Pro at any point in this cycle, so there is no public number to compare against GPT-5.6 Sol or Muse Spark 1.1. That absence is itself informative: Google is choosing silence over a weak number.
A 4% single-day move on a company Alphabet’s size reflects a market recalibrating its view of Google’s model cadence, not a one-time reaction to a delay alone. Google I/O created an expectation in May; missing it in July signals the gap between Google’s internal roadmap and its external promises is wider than investors assumed. The company’s mention of being “productively engaged with the U.S. government” suggests regulatory or public-sector testing may also be shaping the release calendar, though Google did not elaborate on how.
For operators building on Gemini today, the practical read is that Google’s current coding models are the ones you get for the foreseeable near term, not the upgrade promised in May. Teams evaluating a switch to GPT-5.6 Sol or Muse Spark 1.1 for coding-heavy agentic workflows should benchmark those options now rather than wait on a Gemini 3.5 Pro release date Google has stopped committing to.
CNBC (Jonathan Vanian) reported this July 16, 2026, citing Bloomberg’s original reporting on the delay.
