Optimizely has introduced a new family of purpose-built AI models for marketing. The models aim to deliver frontier-level quality at significantly lower costs.
Marketing-specific models trained by the company have been designed to perform marketing tasks and not other general ones. During the early testing phases, they were 10 times more cost-efficient than the latest LLMs. While other models are frontier models, the marketing AI models created by Optimizely are domain-specific. For this reason, they don’t use unnecessary parameters and context.
The marketing team uses many generalized models for multiple functions. This is because the functions can be categorized into writing campaigns, analyzing experiments, and understanding customer behavior. Nevertheless, knowing the function is different from knowing the brand. Optimizely created its models keeping in mind the context of efficiency in marketing operations.
A global survey by the company interviewed more than 2,000 business-to-business marketing executives, and 53% believed the present-day AI is good enough at conveying brand information. But the same survey revealed that the AI had issues relating to the emotional part of marketing. For this reason, Optimizely came up with its own models.
“The initial results of our AI Lab have surpassed our expectations,” said Imran Yousuf, Head of AI Engineering at Optimizely. “We believe post-training models are the long-term solution to the challenging cost vs. quality debate. Our purpose-built family of models is built to get work done fast and cost-effectively, without sacrificing quality. Customers shouldn’t have to think about which model to use, when, or how. They can just trust that our agent platform will apply the right model for the right use case, so their time and their budget go toward the work that actually grows the business.”
Mark-Bench Sets a New Standard for AI Marketing Evaluation
Furthermore, Optimizely created Mark-Bench, which is an open-source benchmark to assess the performance of AI systems in terms of various marketing-based tasks. The Mark-Bench framework assesses AI performance in more than 285 tasks across 15 different marketing domains. It also assesses performance based on over 6,0
Some of these tasks include writing press releases, social media content, and email copy. Therefore, this provides a basis for objectively comparing different AI models and agentic systems. Models used by Optimizely also rely on Mark-IQ, which is the data layer in the Agent Platform. This includes the organizational context, experiment history, and web analytics.
According to Mark-Bench evaluations, Optimizely Agent Platform achieved a 67% all-pass rate. Claude Code scored 60%, while Optimizely reported costs at approximately half the level.
“The problem with generic harnesses is that they’re good for general use and productivity but very inefficient when it comes to domain-specific work,” said Shafqat Islam, President of Optimizely. “Marketing is a unique domain that requires agent harnesses to be specialized. By delivering frontier quality at lower costs, our customers finally have a way to scale agentic marketing to meet real-world demands rather than getting stuck in pilot purgatory.”
This follows the release of Virtual Teammates by Optimizely. The combination of these two will be able to provide special models and specific AI features for marketers. With more agentic workload, choosing the appropriate model will become even more crucial. Optimizely is planning on offering their AI models to all Agent Platform customers.
Want more marketing excellence stories? Visit MarTech News for insights, trends, and expert updates.
News Source: PRNewswire.com
