OpenAI unveils GPT-4 Turbo: How it is better than GPT-4
What's the story
OpenAI has unveiled GPT-4 Turbo, an upgraded version of its leading AI text generator, GPT-4, during its developer conference.
The company asserts that the new model is both more potent and cost-effective than its predecessor.
GPT-4 Turbo comes in two flavors: one for text analysis and another that comprehends both text and image context.
The text analysis model is accessible in preview through an API starting today, with plans to roll out both versions for general availability in upcoming weeks.
Details
GPT-4 Turbo pricing and performance improvements
GPT-4 Turbo's pricing is set at $0.01 per 1,000 input tokens (roughly 750 words) and $0.03 per 1,000 output tokens.
OpenAI has fine-tuned GPT-4 Turbo's performance, enabling it to be offered at a 3x lower price for input tokens and a 2x lower price for output tokens compared to GPT-4.
The cost for the image-processing variant will be determined by the image size. For example, submitting a 1080x1080 pixel image to GPT-4 Turbo will cost $0.00765, as stated by OpenAI.
More
Enhanced knowledge base and context window
GPT-4 Turbo features several enhancements over GPT-4, such as an updated knowledge base and a larger context window.
While GPT-4 was trained on web data until September 2021, GPT-4 Turbo's knowledge extends up to April 2023, providing more accurate responses to recent events.
Additionally, the new model has a 128,000-token context window—four times the size of GPT-4's—which is the biggest context window of any commercially available model, even surpassing Anthropic's Claude 2.
Insights
GPT-4 fine-tuning and token rate limit increase
OpenAI is also initiating an experimental access program for fine-tuning GPT-4.
In contrast to the fine-tuning program for GPT-3.5, GPT-4's predecessor, the new program will include increased supervision and guidance from OpenAI teams due to technical hurdles.
The company has declared that it will double the tokens-per-minute rate limit for all paying GPT-4 customers while keeping the same pricing structure.