OpenAI recently announced GPT-4 Turbo, the Assistants API, and integrated multimodal functionality, coupled with price reductions.
These developments are primarily aimed at broadening access to robust AI for both developers and enterprises.
The upgraded GPT-4 Turbo boasts a context window capacity of 128,000 tokens – equivalent to processing 300 pages of text simultaneously – enhancing its ability to engage in more complex and nuanced dialogues.
Additionally, OpenAI has announced that GPT-4 Turbo will offer a cost-effective pricing model, slashing input token prices by a factor of three and output token costs by half compared to its predecessor.
The Assistants API is a new suite of tools to help developers craft bespoke AI agents tailored to specific tasks. Again, this is designed to foster innovation for developers using OpenAI’s models.
Further, OpenAI has integrated features such as computer vision (CV) and text-to-speech directly into its platform. Notably, the API now provides access to DALL-E 3.
Additionally, OpenAI has initiated a Copyright Shield program to safeguard clients from potential infringement disputes while using the platform’s general features. It means OpenAI will foot the bill for legal costs incurred via copyright issues created by using its models. This aligns with similar offerings from Microsoft and Adobe.
These changes look set to spur on a new wave of GPT-4-powered applications, which could include anything from writing tools to text-to-speech and image recognition tools.
Key points of GPT-4 turbo
Here are some key aspects of GPT-4:
- GPT-4 Turbo is available in two versions: one specializes in text analysis, and the other understands both text and image contexts. The text-only model is currently available via an API in preview, with both versions slated for wider release soon.
- The company has set competitive pricing for GPT-4 Turbo at $0.01 per 1,000 input tokens and $0.03 per 1,000 output tokens, making it three times less expensive for input and two times less for output than its predecessor.
- Pricing for the image-processing capability of GPT-4 Turbo will vary based on image size, with a 1080 x 1080 pixel image costing $0.00765.
- GPT-4 Turbo introduces an extended knowledge base updated up to April 2023 and a quadrupled context window of 128,000 tokens, setting a new standard for commercial AI models.
- OpenAI is rolling out a fine-tuning access program for GPT-4 Turbo with increased support and oversight. Additionally, the token rate limit for paying customers will be doubled without altering the current pricing framework.
ChatGPT users had been noticing that the chatbot had started declaring its knowledge cut-off as April 2023, which had been moved forward to January 2022 from the original September 2021, so it seems like OpenAI has already been updating the model for regular users.
OpenAI’s monetization strategy is still evolving, with analysts keeping a close eye on revenues and whether the company can deliver the cash flows required to fulfill visions of creating artificial general intelligence (AGI) in the future.