Flint: new creative model

Springboards said on April 13 that it had launched Flint in alpha, an artificial intelligence model built to generate a wider range of early-stage ideas for marketers and creative teams. (springboards.ai) The company described Flint as a “divergence model,” meaning it is tuned to spread out possible answers instead of clustering around the most likely one. Springboards said the product is aimed at brainstorming tasks such as campaign territories, strategic angles and creative prompts. (springboards.ai) Springboards announced the launch from Sydney and New York and called Flint an alpha release, signaling an early product stage rather than a finished commercial system. A Globe Newswire release distributed April 13 used the same positioning and said the tool was designed for marketers and creatives. (markets.financialcontent.com) Most large language models are built to predict the most probable next word, which makes them reliable for factual or formatting tasks but often pushes open-ended prompts toward similar answers. Springboards said Flint was built around the opposite goal: maximizing variation in responses during the idea-generation stage. (springboards.ai) That pitch lands as advertising and marketing firms are trying to separate two uses of artificial intelligence: polishing existing work and generating fresh starting points. Springboards said its platform now focuses on “creative variation,” not just efficiency, as more agencies test where general-purpose models help and where they flatten output. (springboards.ai; springboards.ai) Springboards said it measured output diversity by embedding model responses and mapping them in two-dimensional space, then comparing how much “creative territory” different systems covered on advertising tasks. On its homepage, the company says its responses covered 10 to 30 times more creative territory than leading general-purpose large language models, though it did not name those models in the same claim. (springboards.ai) The company has been building a broader case that creativity needs its own benchmarks, not just reasoning scores or accuracy tests. In June 2025, Springboards and industry groups including the American Association of Advertising Agencies, D&AD and The One Club for Creativity announced a project to benchmark large language models on real creative tasks. (springboards.ai; info.springboards.ai) That benchmark material says model rankings shift by brand, culture and task, and that variance matters when teams are looking for inspiration rather than a single correct answer. Flint turns that argument into a product: a model sold on producing more possible directions before humans decide what is worth keeping. (info.springboards.ai; springboards.ai) The launch does not settle whether more variation leads to better campaigns, and Springboards says Flint is still in alpha and lists limitations on its product page. But the company is making a clear bet that, for creative work, the first draft people need from artificial intelligence is not the safest answer. (springboards.ai)