Model family by Ivo · used by 1 application
Sage is Ivo's free, open-source model post-trained via reinforcement learning for long-horizon contract work. It is built on DeepSeek V4 Flash and was post-trained on 1,040 tasks from the Legal Agent Benchmark (LAB), built in partnership with River AI. Released on Hugging Face under an MIT license. In Ivo's LAB testing Sage reached 91% of grading criteria on contract tasks (vs 70% for the DeepSeek V4 Flash base) and an all-criteria pass rate of 15.2% on general tasks, at $0.40-$1.24 per task.
Six dimensions, evidence-linked
Work at Ivo? Claim this listing to correct or complete the data.