Google Launches Veo and Gemini Image Models in OpenAI Compatibility Layer

March 26, 2026

Details

  • Google AI Developers announced that Veo video generation and Gemini image models, including Nano Banana, are now live in the OpenAI compatibility layer.
  • Developers can test these models in existing Python or JavaScript pipelines with zero SDK rewrite, requiring only three code changes: api_key, base_url, and model.
  • The compatibility supports Gemini models via OpenAI libraries (Python, TypeScript/JavaScript) and REST API, using a Gemini API key and base URL like https://generativelanguage.googleapis.com/v1beta/openai/.
  • Key features include image understanding, image generation (e.g., gemini-2.5-flash-image), chat completions, and multimodal capabilities, with code examples provided for encoding images and generating responses.
  • This builds on prior Gemini OpenAI compatibility from November 2024, now extending to Veo video models and advanced image generation like Nano Banana.
  • Official documentation details migration, with Vertex AI supporting Google Cloud Auth.

Impact

Google's expansion of OpenAI API compatibility to Veo and Gemini image models like Nano Banana pressures rivals like OpenAI by enabling seamless migration for developers locked into OpenAI SDKs, potentially accelerating adoption of Google's frontier models. This lowers switching costs in a market where OpenAI holds API dominance, while Gemini's multimodal strengths—such as superior MMMU image understanding scores (70.7% for Flash 2.0 vs. prior versions)—narrow performance gaps. It aligns with intensifying competition, as seen in DeepMind's rapid iterations, fostering broader access to cost-effective video and image generation without ecosystem lock-in.

Read more