Small reasoning model

The talk explains Pleias' new small reasoning models built on open‑source data for RAG, semantic annotation, creative writing, and custom reinforcement learning, with a demo.

Overview

The talk describe the new series of small reasoning model currently developed by Pleias based on the Pleias 1.0 series of base LLM exclusively trained on open data for RAG, semantic data annotation and creative writing. With a dedicated focus on synthetic reasoning data design for mid-training and specialized reinforcement learning with custom reward function. Talk will include a short demo of unreleased models.

Tech stack