New embedding models and API updates

OpenAI announced new embedding models, lower GPT-3.5 Turbo prices, an updated GPT-4 Turbo preview and moderation model, and new API key and usage controls. The text-embedding-3-small model reduces the price per 1k tokens from $0.0001 to $0.00002 and raises the MIRACL average from 31.4% to 44.0% compared with text-embedding-ada-002. The larger text-embedding-3-large model raises the MIRACL average to 54.9%, and both models support shortening embeddings through the dimensions parameter. GPT-3.5 Turbo input prices fall 50% to $0.0005 per 1K tokens, and output prices fall 25% to $0.0015 per 1K tokens.