Dive Brief:
- OpenAI released a smaller, less expensive version of its GPT-4o model Thursday, according to a blog post. The model is more than 60% cheaper than GPT-3.5 Turbo, the company said.
- GPT-4o mini supports text and vision in the API, and OpenAI said the model can call multiple APIs, use full code bases as context and support real-time responses for chatbots. The mini model’s knowledge cutoff is October 2023.
- GPT-4o mini is now available in ChatGPT Free, Plus and Team plans, replacing GPT-3.5. Enterprise users will have access starting next week, OpenAI said.
Dive Insight:
Smaller models have proliferated as organizations look to leverage generative AI capabilities at a lower price. Choosing the right language model for a use case depends on an organization’s needs and available resources.
Microsoft released its Phi-3 family of open, small language models in April. Google unveiled a set of lightweight models in February, called Gemma. Platforms such as Hugging Face offer access to a slew of smaller models, too.
While smaller models can reduce costs, they don’t diminish the risks associated with their larger counterparts. To address concerns, OpenAI tapped a team of specialists to assess GPT-4o mini.
“More than 70 external experts in fields like social psychology and misinformation tested GPT-4o to identify potential risks, which we have addressed and plan to share the details of in the forthcoming GPT-4o system card and preparedness scorecard,” OpenAI said.
Despite generative AI being a strategic enterprise priority, most tech leaders don’t think their business is ready for widescale adoption.
IT pros are worried about the data that will fuel the systems, and enterprise tech infrastructures are not yet up to par. Costs and security concerns also stand in the way of scalable projects.
CIOs are working to level the hurdles that threaten progress, resisting vendor-led pressure for speedy adoption. Across industries, tech leaders are performing organizational readiness assessments and collaborating with C-suite counterparts and stakeholders to set their own pace.
Vendors are looking to sweeten the deal by adding security and safety guardrails. OpenAI launched several compliance and administrative tools for ChatGPT Enterprise Thursday, including APIs to assist with audit requirements, data loss prevention and policy enforcement.