officialmobai_Rostam-S1-4B-GGUF

officialmobai_Rostam-S1-4B-GGUF

GGUF quantizations of Rostam-S1-4B, a 4B reasoning model for math, physics, software work and Persian-English text. For people who run small models locally and want to pick a file size.

Made by bartowski 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: These are GGUF quantizations of Rostam-S1-4B, a reasoning model from Mob AI, made with the Spark project. The model aims at math, physics, software engineering and work in both Persian and English. The files were produced by bartowski with llama.cpp and an importance matrix, in sizes from 2.62 GB to 8.23 GB.

Who it suits: People who run models on their own computers and want a small reasoning model that handles Persian as well as English.

Industry | Who uses it | How they use it | Example
Education | Students and tutors | Work through math and physics questions in Persian or English | A step-by-step explanation of a physics problem
Software | Developers | Ask a local model for coding help and tool-style calls | A function reviewed on a private machine
Language work | Persian-English writers | Move between the two languages in one chat | A question in Persian answered in English
Research | Hobbyists | Compare quant sizes on a small model | Q4_K_M against Q6_K on the same prompt

Skip it if you need image or audio input, because the card lists text input only.

Where to use it:

• Your own computer: any Windows, macOS or Linux machine where Python can fetch the files
• Local runners that read GGUF files, such as llama.cpp, which made these files

Use case | Who | What they get back
Math and physics help | Students | Worked reasoning followed by an answer
Coding help | Developers | Code suggestions, with tool definitions supported in the prompt format
Bilingual chat | Persian and English speakers | Replies in either language

How to set it up: You need Python with pip, enough disk space for the file you choose, and a local runner that opens GGUF files.

• Python and pip on Windows, macOS or Linux
• Disk space for one file, about 2.62 GB for Q4_K_M
• A GGUF-capable runner

1. Install the Hugging Face hub tool.
2. Download the repository's files, or just the one quant you want.

pip install -U huggingface_hub
hf download bartowski/officialmobai_Rostam-S1-4B-GGUF

How to use it: Load a file in your runner and chat with it. The prompt format has a system part, then your message, then a bot turn that opens a thinking channel. Inside the system part you set a thinking effort of low or high.

hf download bartowski/officialmobai_Rostam-S1-4B-GGUF

Tool | What it does
hf download | Copies this repository's files to your computer
thinking_effort | Set to low or high to guide how much the model thinks before it replies
Tool definitions | An optional block in the system part that describes functions the model may call

Example: You set the thinking effort to high and ask in Persian for a step-by-step solution to a physics problem. The model thinks first in its thinking channel, then writes the solution.

For newcomers and experts: If you are new to this, take the Q4_K_M file, which the card names as a good mix of size and speed. Load it and ask one short question.

If you know the field, the files are llama.cpp imatrix quants of a 4B source checkpoint, with BF16 as the full-precision option. The card lists speculative decoding as unsupported. The prompt format uses start and end of sentence markers with user and bot roles.

Choose a quant size
Download the GGUF file
Load it in a local runner
Chat with thinking effort set

Good to know: Larger quants keep more quality but need more disk and memory. The card calls Q6_K and Q5_K_M recommended choices.

Title: File size by quant type in GB
BF16: 8.23 GB
Q8_0: 4.38 GB
Q6_K: 3.61 GB
Q5_K_M: 3.16 GB
Q4_K_L: 2.81 GB
Q4_K_M: 2.62 GB

• License: Apache 2.0
• Cost to run: none named
• Data it reaches: the files you download from Hugging Face, plus the text you type into your own runner
• Not for: input other than text, and speculative decoding, which the card lists as unsupported

On Hugging Face, 14 November 2026

• Task: text generation
• Parameters: 4.1B
• Weights: about 2.1 GB at 4-bit (estimate)
• Formats: GGUF
• Downloads in the last 30 days: 1,091
• Likes: 2
• License: apache-2.0
• Languages: en, fa
• Based on: officialmobai/Rostam-S1-4B
• Tags: gguf, reasoning, math, physics, persian, code, agentic, mob-ai, rostam, text-generation, conversational

Links

• Model: https://huggingface.co/bartowski/officialmobai_Rostam-S1-4B-GGUF
• Files: https://huggingface.co/bartowski/officialmobai_Rostam-S1-4B-GGUF/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download bartowski/officialmobai_Rostam-S1-4B-GGUF

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