K2-Horizon-7B-Uno
Paper: Unlocking Lossless Speedups in LLMs via Discrete Diffusion
Project Page: https://s-sahoo.github.io/uno/
Code: https://github.com/ifm-ai/uno
We present K2-Horizon-7B-Uno, a diffusion-augmented LLM with two pathways in a unified architecture:
- The AR pathway uses the AR weights of the K2-Horizon-7B model.
- The diffusion pathway augments these weights with LoRA-based diffusion adapters.
This repository contains the adapter only. The base-model weights are hosted separately in IFM/K2-Horizon-7B.
Code
Selected evaluation scripts are available in scripts/k2_horizon. The full evaluation suite will be released soon.
Evaluation results
The main number is the benchmark score and the subscript is TPF. Bold marks the
best score in each row. -- and N/A denote unavailable results.
OursReference models
BenchmarkUno
8BNemotron
14BMercury 2
size N/ADiff-Gemma
26B-A4B
Agentic Tool Use
τ3 Banking25.8----9N/A--
τ2 Telecom90.12.4614.34.871N/A68.118.8
τ2 Retail67.13.005.63.1--65.523.7
Terminal-Bench v2.139.12.654.57.527N/A14.714.1
Agentic Coding
SWE-bench Verified70.13.080.81.5--18.75.6
Long-Context Reasoning
AA-LCR68.02.857.31.136N/A19.710.8
Science and Knowledge
Humanity's Last Exam18.62.472.67.216N/A9.214.1
GPQA-Diamond77.12.7140.47.677N/A70.711.9
AA-Omniscience14.32.2311.011.120N/A17.79.9
Math
GSM8K95.42.8893.16.1--95.128.9
MATH50098.92.8689.25.6--92.424.1
AIME-2493.02.9056.74.9--73.716.9
AIME-2590.72.9640.04.5--74.318.9
AIME-2686.32.8446.74.8--70.717.8
Coding
MBPP84.12.4473.85.3--80.115.5
HumanEval95.22.6684.87.5--95.128.2
Efficiency Summary
Average TPF2.715.41N/A17.56
System Throughput5,2552,7941,1971,136
Per-request Throughput405290769836
Citation
If you find this model useful, please cite:
@misc{k2_horizon_7b_uno,
title = {K2-Horizon-7B-Uno},
author = {Institute of Foundation Models},
year = {2026},
howpublished = {\url{https://huggingface.co/IFM/K2-Horizon-7B-Uno}},
}