TokenRhythm/NeoHorse
NeoHorse-1: Towards Recursive Self-Improvement via Agentic Post-Training with Routing Harness.
About the project
The NeoHorse family of open-weight models for agent workflows: NeoHorse-1 (4B and 9B) for tool use, coding, and instruction following, and NeoHorse-Jev-4B which turns application state into structured decisions via prefill-only inference.
Useful for
- Run NeoHorse-1-4B locally in GGUF quantization for agent tasks without cloud
- Use NeoHorse-Jev-4B to pick a tool or route a request based on application state
- Post-train the model for your own harness via routing-guided agentic post-training
README summarized by DeepSeek V4.1 Flash. Details may be inaccurate.
Why it’s trending
- 0 stars so far today, about 44 expected by the end of the day.
- The repository is 21 days old and already has 947 stars.
- Top new repositories this month: #100.
Stars per day
Bars are daily stars, the line is the usual pace. Red marks spike days.
Numbers
- Total stars
- 947
- Today
- 0 · ≈ 44 by evening
- Forks
- 11
- Issues and pull requests
- 2
- Watchers
- 1
- Language
- Python
- License
- Apache-2.0
- Created
- September 4, 2026
- Last push
- September 24, 2026
Star trust
Star growth looks organic. The magnitude is lowered, not zeroed:
- Fewer forks than usual: 11 for 947 stars.
These are heuristics, not a verdict: we judge by the repository’s behavior, not by a list of stargazers.
Spotted in
- September 25, 2026Top new repositories this month: #100
Share
README badge
Paste it into your README — the badge shows the current magnitude and links to this page.
[](https://gitnova.dev/en/r/TokenRhythm/NeoHorse)More in this category
-
6.5
jaredpalmer/kev
kev is a LoRA adapter with a small readout head on top of Qwen2.5-0.5B that answers many typed questions about a document in a single forward pass, returning calibrated probabilities instead of text.
-
5.9
deepopen-com/deepopen
Open-source non-autoregressive System 1 decision engine for typed decisions: returns typed answers (choice, score, noul) in a single forward pass across 100+ languages without generating text. Suited for…
-
5.7
togethercomputer/tev1
Open dataset and fine-tuning recipe for Qwen3.5-4B that picks one of 2–24 options given context and a question. The repo ships data, build scripts, Together AI training and evaluation tooling.
-
5.2
NVIDIA/Model-Optimizer
NVIDIA library for compressing and accelerating models: quantization, pruning, distillation, NAS and speculative decoding with export to TensorRT-LLM, vLLM, SGLang. For ML engineers preparing models for deployment.
-
4.7
Liuziyu77/Valen
A multimodal decision model built on a Qwen3.5-0.8B/2B backbone: it takes text, images and video with an instruction and returns probabilities over supplied candidates without generating answer tokens. The repo…
-
4.5
mizorewww/laya-mlx
Native MLX runtime for Laya typed decision models: returns probabilities for choices, scores or truth values without text generation, running locally on Apple Silicon.