Abacus.AI launched Smaug Agentic, Smaug Flash and Smaug Mini, a family of open-weight language models built for enterprise AI agents. The company frames them around cost, claiming large reductions in the price of running agent workloads compared with calling the largest commercial models.
Open-weight releases let a business run a model on its own infrastructure rather than paying per call to an outside service, which matters when an agent makes thousands of calls to finish one job.
What it means: for high-volume, repetitive tasks the cheapest capable model usually wins, not the most powerful one. Open-weight options give smaller companies a way to control both cost and where their data goes, instead of renting intelligence by the token.