K2-Horizon-0.9B

K2-Horizon-0.9B

Answers maths, coding and science questions as a compact open reasoning model that reads up to 128K tokens at once. Built by IFM for people who want a small model with public data and code.

Made by IFM on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: K2-Horizon-0.9B is the small dense model in the K2-Horizon family from IFM. It is a decoder-only model in the 0.9B class, tuned for reasoning, with a 128K context window. Its training data, recipe and code are all public.

Who it suits: It suits people who want a compact reasoning model they can inspect, and researchers who want to see exactly how a small model was trained.

Industry | Who uses it | How they use it | Example
Research | Language model researchers | Study a small model whose data and code are public | Read the training recipe and run the public evaluation code
Software | Developers | Try a compact reasoning model on maths and code questions | Send it a coding question and review the answer
Education | Students and teachers | Learn how a model is trained from open material | Follow the public training recipe step by step

People who need the depth of a much larger model can skip it, and so can anyone who needs languages other than English and Chinese.

Where to use it: The card excerpt names no app. The files load in any Python project that can read them.

• Windows, macOS and Linux, with Python for the download
• Python projects built on the transformers library, which the record lists

Use case | Who | What they get back
Maths and coding questions | Students and developers | A written answer from a small reasoning model
Long inputs | Anyone with long texts | A reply that can draw on up to 131,072 tokens of context

How to set it up: You need Python with pip, and room for the weights.

• Python with pip
• About 2.2 GB of disk for the BF16 weights, or about 539 MB at 4-bit, which is an estimate

1. Install the Hugging Face hub library with pip.
2. Download the repository's files to your computer with the hf command.
3. Load the downloaded files with the transformers library in your own script.

pip install -U huggingface_hub
hf download IFM/K2-Horizon-0.9B

How to use it: The card excerpt shows no run command, so use the transformers library the record lists. Load the downloaded model and its tokenizer, send a prompt and read the reply. Its window reaches 131,072 tokens, and the card names YaRN RoPE scaling as the way it gets there.

Example: Ask it a short maths or coding question, such as a small algebra problem or a request to explain a function. The card says the model was tested on maths, coding, science and tool-use benchmarks, so those are the jobs it was built for.

For newcomers and experts: If you are new to this, download the files with the two commands above and load them with transformers. Try one short maths question first. Hugging Face lists 1.1B parameters, while the card calls the model 0.9B class.

If you know the field, this is a dense decoder-only model with YaRN RoPE scaling for long context. It was trained by multi-teacher distillation, with separate teachers for maths and code, STEM and instruction following. A model this small trades breadth for a small download.

You send a prompt
The dense decoder-only model reads it, up to 128K tokens
The model writes its reasoning and answer

Good to know: The card's comparison table lists the parameter counts below next to other small models.

• License: Apache 2.0, as the Hugging Face record states
• Languages: English and Chinese in the record
• Other versions: the record counts 16 quantized copies and 11 fine-tuned ones made by other people
• Cost to run: none named
• Data it reaches: the model files you download from Hugging Face, plus the prompts you send it

Title: Parameters in the card's comparison table, in billions
K2-Horizon-0.9B: 0.9 billion
Qwen3.5-0.8B: 0.8 billion
OpenBMB-1B: 1 billion
Qwen3.5-2B: 2 billion

On Hugging Face, 13 November 2026

• Task: text generation
• Library: transformers
• Parameters: 1.1B
• Weights: about 2.2 GB in BF16; about 539 MB at 4-bit (estimate)
• Formats: Safetensors
• Downloads in the last 30 days: 11,244
• Likes: 105
• License: apache-2.0
• Languages: en, zh
• Quantized versions: 16
• Fine-tunes: 11
• Tags: transformers, safetensors, k2_horizon, text-generation, k2-horizon, 0.9b, dense, reasoning, knowledge-distillation, conversational

Links

• Model: https://huggingface.co/IFM/K2-Horizon-0.9B
• Files: https://huggingface.co/IFM/K2-Horizon-0.9B/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download IFM/K2-Horizon-0.9B

Then follow the usage in its model card. Nothing in this kit runs by itself.