allspark releases iris-mini and iris-pro search agents
allspark team releases two open-source search agents, iris-mini with 35b parameters and iris-pro with 397b parameters, outperforming others in their class. these agents impact the development of autonomous web research capabilities.
the allspark team released two open-source search agents, iris-mini with 35 billion parameters and iris-pro with 397 billion parameters. both models build on qwen-series models and work with a 256,000-token context window. they outperform other open-weight search agents in their respective size classes. the training data and models also improved performance on tasks they were never trained for, including general tool use and office work. this release impacts the development of autonomous web research capabilities.