Gemini Pro delay leaves Google with an empty flagship tier
Polymarket, the crypto-based prediction platform, puts the probability of no new Gemini Pro release before 30 September at 69%.
Every specific launch date on the market trades in low single digits, with 28 September the next most favoured outcome at 8%.
The bet centres on Gemini 3.5 Pro, the model Sundar Pichai promised developers would arrive within a month at Google's I/O conference on 19 May.
That deadline has come and gone three times over. The model missed targets in June, mid-July and early August, and as of late August still has no model identifier, no pricing and no launch date.
Google DeepMind's own models page continues to list it as coming soon. The hold-up is technical. Google has fallen months behind because it has been trying to improve the model's capabilities, particularly in coding.
The company reset and updated its training data in late June to fix the code-generation problems, only for internal results to disappoint again.
The delay carries a competitive cost. Google engineers and researchers are reported to be concerned that the company risks losing its edge as Anthropic and OpenAI ship models that outperform Gemini.
Rivals have not stood still. OpenAI launched its GPT-5.6 family in July, and Elon Musk's Grok 4.5 opened to the public in the same window.
Anthropic, meanwhile, has rolled out its Fable 5 model. Rather than rush the Pro release, Google has filled the gap with cheaper, faster models.
It has shipped a run of Flash models, with Gemini 3.7 Flash reaching general availability on 13 August.
The awkward result is that the strongest Gemini model users can currently call is not a Pro model at all.
Google has also confirmed it has begun pre-training an entirely new flagship, Gemini 4. That has fuelled speculation that the company may effectively skip ahead rather than salvage the delayed release.
Traders appear unconvinced a fix is imminent. A related Polymarket contract puts the chance of no Pro model by 1 November at roughly even.
