Instructions to use SirSahOl/opt-125m-chat-mlx-8bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use SirSahOl/opt-125m-chat-mlx-8bit with MLX:

# Make sure mlx-lm is installed
# pip install --upgrade mlx-lm
# if on a CUDA device, also pip install mlx[cuda]

# Generate text with mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("SirSahOl/opt-125m-chat-mlx-8bit")

prompt = "Once upon a time in"
text = generate(model, tokenizer, prompt=prompt, verbose=True)

Transformers

How to use SirSahOl/opt-125m-chat-mlx-8bit with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="SirSahOl/opt-125m-chat-mlx-8bit")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("SirSahOl/opt-125m-chat-mlx-8bit")
model = AutoModelForCausalLM.from_pretrained("SirSahOl/opt-125m-chat-mlx-8bit", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings
LM Studio

vLLM

How to use SirSahOl/opt-125m-chat-mlx-8bit with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "SirSahOl/opt-125m-chat-mlx-8bit"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "SirSahOl/opt-125m-chat-mlx-8bit",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/SirSahOl/opt-125m-chat-mlx-8bit

SGLang

How to use SirSahOl/opt-125m-chat-mlx-8bit with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "SirSahOl/opt-125m-chat-mlx-8bit" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "SirSahOl/opt-125m-chat-mlx-8bit",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "SirSahOl/opt-125m-chat-mlx-8bit" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "SirSahOl/opt-125m-chat-mlx-8bit",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

MLX LM

How to use SirSahOl/opt-125m-chat-mlx-8bit with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Generate some text
mlx_lm.generate --model "SirSahOl/opt-125m-chat-mlx-8bit" --prompt "Once upon a time"

Docker Model Runner
How to use SirSahOl/opt-125m-chat-mlx-8bit with Docker Model Runner:
```
docker model run hf.co/SirSahOl/opt-125m-chat-mlx-8bit
```
Atomic Chat

opt-125m-mlx-8bit

8-bit MLX conversion of facebook/opt-125m optimized for Apple Silicon native GPU inference.

Converted by: SirSahOl
Source Model: facebook/opt-125m
Framework: MLX by Apple
Quantization: 8-bit (Average 8.25 bits per weight)
Format: safetensors
License: other

Model Details

Architecture: OPTForCausalLM
Parameters: 125M
Context Length: 2,048 tokens
Format: MLX (Apple Silicon native GPU format)
Quantization: 8-bit (Average 8.25 bits per weight)
Active VRAM Footprint: ~214 MB (Minimum recommended: 8 GB Unified Memory)

Quick Start

Installation

pip install mlx-lm

Usage

CLI

# Chat interactively
mlx_lm.chat --model SirSahOl/opt-125m-chat-mlx-8bit

# Generate text
mlx_lm.generate --model SirSahOl/opt-125m-chat-mlx-8bit --prompt "Write a short poem about artificial intelligence."

Python API (with Chat Template)

from mlx_lm import load, generate

model, tokenizer = load("SirSahOl/opt-125m-chat-mlx-8bit")

messages = [
    {"role": "user", "content": "Explain quantum superposition in simple terms."}
]
prompt = tokenizer.apply_chat_template(
    messages, tokenize=False, add_generation_prompt=True
)

response = generate(model, tokenizer, prompt=prompt, verbose=True)
print(response)

Performance Benchmarks

Measured Benchmarks (Apple M1)

Metric	4-bit	8-bit	16-bit
Tokens/sec	360.69	264.13	60.9
TTFT	2.79 ms	3.8 ms	16.44 ms
Peak Memory	234.4 MB	335.8 MB	316.7 MB

Benchmarked on Apple M1 with 8GB unified memory. Average over 5 runs with 256 max tokens.

Multi-Quantization Comparison

Evaluate your hardware budget and choose the optimal precision:

Variant	Disk Size	VRAM Footprint	Target Apple Silicon Hardware	Key Advantage
4-bit MLX	~79 MB	~119 MB	M1 / M2 / M3 / M4 (8GB+ Unified Memory)	Ultra-compact footprint, maximum generation speed; negligible memory pressure.
8-bit MLX (This Repository)	~146 MB	~191 MB	M1 / M2 / M3 / M4 (8GB+ Unified Memory)	Near-lossless precision with an extremely lightweight footprint.
16-bit MLX	~270 MB	~340 MB	M1 / M2 / M3 / M4 (8GB+ Unified Memory)	Full unquantized precision; zero perplexity penalty for reference evaluation.

Who Should Use This?

Your Hardware	Recommended Quantization
M1/M2/M3/M4 (8GB – 16GB)	4-bit — Best balance of speed, low memory, and multitasking capability
M1/M2/M3/M4 Pro/Max (18GB – 36GB)	8-bit — Higher quality reasoning with comfortable memory headroom
M1/M2/M3/M4 Max/Ultra (36GB – 192GB)	16-bit — Unquantized full precision, zero quality degradation

General guidance:

Use 4-bit if you want to run this model alongside IDEs, browsers, and background development tools.
Use 8-bit if you have 16GB+ unified memory and require superior reasoning and code accuracy.
Use 16-bit for research, benchmarking, evaluation, or high-end workstation deployments.

Other Quantization Variants

Variant	Link
4-bit	SirSahOl/opt-125m-chat-mlx-4bit
8-bit	SirSahOl/opt-125m-chat-mlx-8bit
16-bit	SirSahOl/opt-125m-chat-mlx-16bit

LM Studio & Local Inference Setup Guide

To prevent runaway loops and ensure correct conversational turn-taking, configure custom stop tokens in your local inference runtime:

<|im_start|>
<|im_end|>
<|endoftext|>

Prompt Template Formatting

System Prefix: <|im_start|>system\n
System Suffix: <|im_end|>\n
User Prefix: <|im_start|>user\n
Assistant Suffix: <|im_end|>\n<|im_start|>assistant\n

Ollama Quickstart

FROM SirSahOl/opt-125m-chat-mlx-8bit
PARAMETER stop "<|im_start|>"
PARAMETER stop "<|im_end|>"
PARAMETER stop "<|endoftext|>"
PARAMETER temperature 0.7

ollama create opt-125m-chat-mlx-8bit -f Modelfile
ollama run opt-125m-chat-mlx-8bit

Conversion Details

Limitations & Known Issues

4-bit group-wise quantization introduces minor precision loss compared to unquantized weights; for deep mathematical derivations or precision-critical reasoning, test the 8-bit or 16-bit variants.
High context sequences (>32K tokens) require sufficient unified memory headroom; ensure unified memory is not overcommitted.
This is a weight-only MLX conversion designed specifically for Apple Silicon GPUs (M1/M2/M3/M4 series).

License

This model conversion inherits the license of the source model: other.

See the original model card for full license details.

Changelog

Version	Date	Changes
v1.0	N/A	Initial conversion

Converted with MLX Foundry — a professional pipeline for converting models to Apple MLX format.

Downloads last month: -

Safetensors

Model size

0.2B params

Tensor type

U32

F16

MLX

Hardware compatibility

8-bit

Model tree for SirSahOl/opt-125m-chat-mlx-8bit

Base model

facebook/opt-125m

Quantized

(13)

this model

Collection including SirSahOl/opt-125m-chat-mlx-8bit

MLX Models by SirSahOl — Optimized for Apple Silicon

Collection

Curated collection of high-performance MLX model conversions benchmarked and optimized for Apple Silicon (M1/M2/M3/M4 unified memory). • 74 items • Updated about 12 hours ago