Vertex AI
$300 in free credits to get started
Models for text generation and embeddings are available (since June 7, 2023). Usage is done via an API in a way quite similar to what OpenAI offers. https://cloud.google.com/vertex-ai/docs/generative-ai/start/quickstarts/api-quickstart
For embeddings, the API accepts 3072 tokens as input and produces vectors of 768 dimensions. At Google, the pricing is per character, at $0.0001, or 0.01c for 1000 characters. Reminder: OpenAI charges 0.4c per 1K tokens, so Google is slightly cheaper.
For text generation, the price is 0.02 for 1000 tokens (for Davinci, it's 0.03 for GPT-4 in prompt and 0.06 in output). Google is therefore a bit cheaper (depending on the average length of a token).
The performance of Google's models, especially in terms of formal reasoning, is considered less good than that of OpenAI, which justifies the difference.
Google has specialized models for code completion that OpenAI does not have. The price for 1000 characters for code completion is: $0.0005.
In general, Google's services seem less expensive and are better for specific tasks, but not as good for very complex tasks like logical reasoning or tasks requiring very large prompts (considering OpenAI offers prompts with sizes up to 32,000 tokens).