Google releases Gemini 1.5 Pro in public preview on Vertex AI, offering support for up to 1M tokens, and says the model will power new features in Code Assist
Gemini 1.5 Pro, Google's most capable generative AI model, is now available in public preview on Vertex AI, Google's enterprise-focused AI development platform.
Context & Ripple Effects
Google had already positioned Gemini 1.5 for developers and enterprise users around a 1M-token context window, while Vertex AI had gained adapter-based tuning and related developer tooling in the Gemini 1.0 rollout. This preview turns that model capability into an enterprise-platform offering rather than a standalone model announcement.
The addition of Gemini-powered Code Assist features ties the model’s long-context capacity to a concrete developer workflow, extending Google’s effort to make Vertex AI a delivery layer for its model portfolio.
First-order effects
- Vertex AI customers can test Gemini 1.5 Pro with up to 1M tokens of context, giving enterprise developers access to the model through Google’s managed AI platform.
- Google can use Gemini 1.5 Pro as the foundation for new Code Assist capabilities, connecting model availability to its developer-tool product line.
Second-order effects
- Long-context support becomes a more visible selection criterion for enterprise AI platforms, increasing pressure on competing model and cloud offerings to match both capacity and managed deployment.
- Customers evaluating coding assistants may increasingly assess the surrounding cloud and AI platform—not just the assistant—because model access, tuning tools, and developer workflows are being bundled together.
Third-order effects
- If this pattern persists, frontier models will be differentiated less by isolated releases and more by how tightly they are integrated into enterprise AI platforms and daily software-development tools.
- Very large context windows could become a standard enterprise feature, shifting competition toward the cost, reliability, and governance of processing larger proprietary inputs rather than context length alone.
The trend: This is one data point in the platformization of frontier AI, where cloud providers commercialize model advances through integrated enterprise and developer products.