According to a new report from The Information OpenAI Scheduled for release in the next two weeks Strawberry(Strawberry)As its ChatGPT part of the service, earlier than previously expected in the fall.
While Strawberry is part of ChatGPT, it is a standalone product, and one possibility would be to include Strawberry in the drop-down menu of AI models that customers can select to support ChatGPT.
The biggest difference between Strawberry and other conversational AIs is its ability to "think" before responding, rather than answering a question immediately, with the thinking phase typically lasting 10-20 seconds.
The initial version is said to only be able to receive and generate text, not images, meaning it is not yet multimodal like other OpenAI models, and the decision to release it in plain text may reflect the pressure OpenAI feels to release products as it faces more competition.
Regarding pricing, Strawberry may differ from existing ChatGPT, which has free and subscription pricing tiers.
It may limit the maximum number of messages a user can send per hour and may offer higher priced tiers for faster response times, according to a person with knowledge of the product.
In early internal discussions at OpenAI, subscription prices have been as high as $2,000 per month, but have not yet been finalized.ChatGPT expects ARR to reach $2 billion in revenues, primarily from the $20 per month subscription fee, with the cost also including free packages used by hundreds of millions of people each month.
In addition, some users have also discovered that ChatGPT has uploaded a ChatGPT Pro subscription, which costs $200 per month.
More advanced models such as Strawberry and Orion can be more expensive to train and run than previous models. Given more time to think, the Strawberry model can answer more complex questions than OpenAI's current models.
Strawberry's extra thinking or processing time may mean more computational power, and Strawberry is also expected to be easier to use than GPT-4o for complex or multi-step queries, and therefore more costly, which OpenAI may want to pass on some of the cost to customers.
Currently, customers must enter a variety of additional words into ChatGPT to get the answers they want, such as telling the chatbot to complete intermediate reasoning steps to arrive at the final answer, also known as "thought chain prompts.
Strawberry's features are designed to help customers avoid this or other hacks to get smarter results. This means that Strawberry is not only better at math problems and coding, but also at more "subjective" business tasks, such as brainstorming product marketing strategies.
In such tasks, Strawberry will provide more company-specific and detailed advice, such as generating a weekly execution plan. According to industry insiders, Strawberry's think phase helps avoid mistakes.
Additionally, the extra time makes it more likely that Strawberry will know when it needs to ask a customer a follow-up question so it knows how to fully answer their question. In theory Strawberry is able to skip the thinking step when people ask simple questions, but in practice it doesn't always do so.
As a result, Strawberry may mistakenly think too long to answer questions that OpenAI's other models can answer in a short period of time.
Some people who have used the Strawberry prototype have complained that it is slightly more responsive than OpenAI's currently released GPT-4o, and that it is not worth waiting the extra 10-20 seconds.
While Strawberry also aims to remember and integrate previous chats with customers before answering new questions - an important detail when users have specific preferences, such as a certain format in which they want to write software code .