NeoHorse
NeoHorse is a family of open-weight language models sized for agent workflows, with 4B and 9B versions tuned for tool use, coding and instruction following, plus a companion decision model that uses prefill-only inference to pick an action or score a condition.
NeoHorse is TokenRhythm's family of open-weight models for agent workflows. NeoHorse-1 provides 4-billion and 9-billion parameter language models tuned for tool use, coding and instruction following, released under the Apache 2.0 license with quantized and MLX builds for running on ordinary hardware.
NeoHorse-Jev builds on NeoHorse-1-4B to make structured decisions through prefill-only inference: an application defines a question and its possible answers, and the model selects a choice, checks a yes-or-no condition, or assigns a rating, which the application then uses to route requests, select tools or control a workflow. The project reports benchmark results comparing NeoHorse-Jev-4B against other open-weight decision models on text-classification and multiple-choice benchmarks.