OpenAI PublishedGPT-4-Turbo official version, this is a model with vision capabilities, capable of handling 128k context. This model is now fully open and can be used through "gpt-4-turbo",up to dateThe version is "gpt-4-turbo-2024-04-09".
Interface information: https://platform.openai.com/docs/models/continuous-model-upgrades
Pricing information: https://openai.com/pricing
Related limits: https://platform.openai.com/docs/guides/rate-limits/usage-tiers?context=tier-five
According to the official statement, the basic capabilities of this model have been significantly improved. It has its own visual capabilities and does not require the use of a 4v interface. Visual requests can now also use JSON mode and function calls. In addition, the training data is available until December 2023.
In terms of price, the price of GPT-4-Turbo remains the same as the previous GPT-4-Turbo.
- Input: $10.00/1 million tokens
- Output: $30.00/1 million tokens
- Read the picture:lowest $0.00085/Figure
In terms of frequency limitation,HighestTake Tire5 as an example,HighestConcurrency is 10,000 times/minute,HighestThe processing rate is 1,500,000 tokens/minute.
OpenAI also demonstrated several use cases using GPT-4-Turbo with vision capabilities. For example, Devin, built by @cognition_labs, is an AI software engineering assistant powered by GPT-4Turbo that can perform various coding tasks using vision.
Another example is the @healthifyme team using GPT-4Turbo with Vision to build Snap, which can recognize various food photos from around the world and provide nutritional insights to users. Finally, Make Real, developed by @tldraw, allows users to draw UIs on a whiteboard and design websites using GPT-4Turbo with Vision to generate code directly.
Overall, GPT-4-Turbo is a powerful model, and its release will bring new possibilities to the field of AI.