Math GPT-OSS Model (13 Experts)

Project: https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/

👥 Follow the Authors

Aman Priyanshu LinkedIn Twitter Website

Supriti Vijay LinkedIn Twitter Website

Introduction

This is a pruned variant of OpenAI's GPT-OSS-20B model, reduced to 13 experts per layer based on activation patterns from the AmanPriyanshu/GPT-OSS-20B MoE Expert Activations dataset. We analyzed router decisions across evaluation benchmarks to identify and retain experts most relevant for math tasks.

⚠️ Experimental Model: This is an experimental pruned model that may not work well - check the examples below to see if the outputs meet your needs before use.

This pruning approach reduces the model size while attempting to preserve performance on the target domain.

Model Architecture & Statistics

Metric Value
Base Model openai/gpt-oss-20b
Architecture Mixture-of-Experts Transformer
Total Parameters ~9.6B (pruned from 21B)
Original Experts per Layer 32
Pruned Experts per Layer 13
Layers 24
Top-k Routing 4
Context Length 128K tokens
Attention Heads 64 (Query), 8 (Key-Value)
Residual Dimension 2880
Attention Pattern Alternating dense & sliding window (128 tokens)
Positional Encoding RoPE (Rotary Position Embedding)
Normalization RMSNorm
Precision BF16
License Apache 2.0
Specialization Math

Pruning Methodology

What is Expert Pruning?

Mixture-of-Experts models contain multiple specialized sub-networks (experts) per layer. During inference, only a subset of experts are activated for each token. Expert pruning involves:

  1. Analyzing Usage Patterns: Tracking which experts activate most frequently for specific tasks
  2. Removing Underutilized Experts: Discarding experts with low activation rates for the target domain
  3. Preserving Router Functionality: Maintaining the routing mechanism with fewer available experts

Our Approach

  • Data-Driven Selection: Used activation patterns from math evaluation tasks
  • Systematic Reduction: Reduced from 32 to 13 experts per layer
  • No Retraining: Direct removal without additional training steps

Performance & Applications

Pruning Benefits

  • Smaller Memory Footprint: 40.6% of original expert parameters
  • Reduced Computational Load: Fewer routing decisions during inference
  • Focused Capabilities: Retains experts relevant to math tasks

Use Cases

  • Speculative Decoding: Draft model for full GPT-OSS-20B
  • Resource-Constrained Deployment: Edge devices, mobile applications
  • Research: Study expert specialization in MoE models
  • Fine-tuning: Smaller base model for domain adaptation

Note: Performance may vary depending on how well the pruned experts match your specific use case.

Motivation & Expert Selection

This mathematics-focused model utilizes experts that exhibited strong performance on mathematical reasoning tasks from MMLU mathematics subjects and quantitative sections. These experts excel at mathematical computation, proof strategies, and logical reasoning.

The expert selection process utilized our comprehensive analysis of router activation patterns across multiple evaluation benchmarks:

  • GPQA: Graduate-level questions in physics, chemistry, biology (Diamond & Expert subsets)
  • MMLU/MMLU-Pro: Comprehensive knowledge across 57+ subjects including science, medicine, law
  • SORRY-Bench: Safety evaluation across harmful content categories
  • Tulu3: Persona-driven instruction following with verifiable constraints
  • Polyglot-or-Not: Multilingual factual completion tasks

By identifying experts that consistently activated for math tasks, we created this specialized model that maintains domain expertise while significantly reducing computational requirements from 32 to 13 experts per layer.

Dataset & Analysis Foundation

This model is based on analysis from the GPT-OSS-20B MoE Expert Activations dataset available at: 🔗 https://huggingface.co/datasets/AmanPriyanshu/GPT-OSS-20B-MoE-expert-activations

The dataset contains router activation patterns from OpenAI's GPT-OSS-20B model across diverse evaluation benchmarks, enabling the creation of these domain-optimized models through systematic expert pruning.

Pruning Methodology

Our approach involves:

  1. Activation Analysis: Comprehensive evaluation of expert usage patterns across domain-specific tasks
  2. Expert Ranking: Identification of the most frequently activated experts for target domains
  3. Systematic Pruning: Reduction from 32 to 13 experts while preserving router functionality
  4. Quality Validation: Testing to ensure maintained performance on target tasks

This is a direct pruning approach - no additional training was performed. The model inherits all capabilities from the original GPT-OSS-20B with focused expert selection.

Usage

CPU Inference

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load the specialized model on CPU
model = AutoModelForCausalLM.from_pretrained(
    "AmanPriyanshu/gpt-oss-9.6b-specialized-math-pruned-moe-only-13-experts", 
    torch_dtype=torch.bfloat16, 
    device_map="cpu", 
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-9.6b-specialized-math-pruned-moe-only-13-experts")

# Generate with the model
messages = [
    {"role": "user", "content": "Solve this equation: 2x + 5 = 17. Show your work step by step."}
]

inputs = tokenizer.apply_chat_template(
    messages, 
    add_generation_prompt=True, 
    return_tensors="pt", 
    return_dict=True,
    reasoning_effort="medium"
)

# Ensure inputs are on the same device as model
inputs = {k: v.to(model.device) for k, v in inputs.items()}

outputs = model.generate(
    **inputs, 
    max_new_tokens=512,
    do_sample=True,
    temperature=0.1,
    top_p=0.9,
    pad_token_id=tokenizer.eos_token_id,
    eos_token_id=tokenizer.eos_token_id
)

# Decode only the generated part
input_length = inputs['input_ids'].shape[1]
response_tokens = outputs[0][input_length:]
response = tokenizer.decode(response_tokens, skip_special_tokens=True)
print(response)

Apple Silicon (MPS) Inference

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Check MPS availability and load model
device = "mps" if torch.backends.mps.is_available() else "cpu"

model = AutoModelForCausalLM.from_pretrained(
    "AmanPriyanshu/gpt-oss-9.6b-specialized-math-pruned-moe-only-13-experts", 
    torch_dtype=torch.float16,  # Better MPS compatibility
    device_map=device, 
    trust_remote_code=True,
    low_cpu_mem_usage=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-9.6b-specialized-math-pruned-moe-only-13-experts")

# Generate with the model
messages = [
    {"role": "user", "content": "Solve this equation: 2x + 5 = 17. Show your work step by step."}
]

inputs = tokenizer.apply_chat_template(
    messages, 
    add_generation_prompt=True, 
    return_tensors="pt", 
    return_dict=True,
    reasoning_effort="medium"
)

# Move inputs to model device
inputs = {k: v.to(model.device) if isinstance(v, torch.Tensor) else v for k, v in inputs.items()}

# Use torch.no_grad for MPS stability
with torch.no_grad():
    outputs = model.generate(
        **inputs, 
        max_new_tokens=512,
        do_sample=True,
        temperature=0.1,
        top_p=0.9,
        pad_token_id=tokenizer.eos_token_id,
        eos_token_id=tokenizer.eos_token_id,
        use_cache=True
    )

# Decode only the generated part
input_length = inputs['input_ids'].shape[1]
response_tokens = outputs[0][input_length:]
response = tokenizer.decode(response_tokens, skip_special_tokens=True)
print(response)

GPU Inference

For CUDA GPU inference, the models work with minimal modifications from the CPU example above. Simply change:

device_map="auto"  # Will automatically use GPU if available
torch_dtype=torch.bfloat16  # or torch.float16

Model Examples

The following examples demonstrate the model's performance on math tasks:

Example 1

Question: Solve this equation: 2x + 5 = 17. Show your work step by step.

Response:

<|channel|>analysis<|message|>Need to solve linear equation. Provide steps.<|end|><|start|>assistant<|channel|>final<|message|>**Step‑by‑step solution**

We are given the equation  

\[
2x + 5 = 17
\]

and we want to find the value of \(x\).

---

### 1. **Isolate the variable term**

The variable \(x\) appears in the first term on the left side.  
To get \(x\) alone, we must remove the constant \(+5\) from the left side.

Subtract \(5\) from both sides:

\[
2x + 5 - 5 = 17 - 5
\]

Simplify:

\[
2x = 12
\]

---

### 2. **Solve for \(x\)**

Now \(x\) is multiplied by \(2\).  
Divide both sides by \(2\):

\[
\frac{2x}{2} = \frac{12}{2}
\]

Simplify:

\[
x = 6
\]

---

### 3. **Check the solution**

Plug \(x = 6\) back into the original equation:

\[
2(6) + 5 = 12 + 5 = 17
\]

The left side equals the right side, so the solution is correct.

---

\[
\boxed{x = 6}
\]<|return|>

Example 2

Question: What is the Pythagorean theorem and how is it used?

Response:

<|channel|>analysis<|message|>Need explain theorem, formula, usage.<|end|><|start|>assistant<|channel|>final<|message|>**The Pythagorean Theorem**  
> In a right‑angled triangle, the square of the length of the hyp‑ (the longest side) equals the sum of the squares of the other two sides.  
> **Formula**:  
> \[
> a^{2}+b^{2}=c^{2}
> \]  
> where \(a\) and \(b\) are the lengths of the legs, and \(c\) is the length of the hyp‑.

---

### How It Is Used

| Field | Practical Use |
|-------|--------------|
| **Geometry & Trometry** | Finding side lengths in a right‑angled triangle. |
| **Engineering** | Calculating dimensions of a bridge, road, or building. |
| **Physics** | Deriving the relationship between velocity, acceleration, and time (e.g., in a 2‑D space). |
| **Computer Graphics** | Rendering 3‑D shapes, calculating distances. |
| **Mat Education** | Teaching basic geometry, problem‑solving. |
| **Everyday Life** | Measuring a piece of a right‑angled object (e.g., a piece of a table). |

---

### Quick Example

Supp you have a right‑angled triangle with legs 3 m and 4 m.  
Using the theorem:  

\[
3^{2}+4^{2}=9+16=25
\]

So the hyp‑ is \(\sqrt{25}=5\) m.

---

**Key Take‑in**:  
The theorem lets us *solve for any side* of a right‑angled triangle, and it’s the foundation of many geometry, physics, and engineering calculations.<|return|>

Example 3

Question: Calculate the area of a circle with radius 7 meters.

Response:

<|channel|>analysis<|message|>Need area = π r^2 = π*49 ≈ 153.938.<|end|><|start|>assistant<|channel|>final<|message|>The area of a circle is given by  

\[
A = \pi r^{2}
\]

where \(r\) is the radius.

For a radius of \(7\) meters:

\[
A = \pi (7\,\text{m})^{2}
   = \pi \times 49\,\text{m}^{2}
   \approx 3.14159 \times 49\,\text{m}^{2}
   \approx 153.94\,\text{m}^{2}
\]

So the area of a circle with a radius of 7 m is about **\(1.54 \times 10^{2}\) m²** (≈ 153.9 m²).<|return|>

Citation

If you use this model in your research, please cite:

@misc{priyanshu2025gptoss,
  title={{GPT-OSS MoE Expert Fingerprinting: Analyzing Expert Activation Patterns in Mixture of Experts Models}},
  author={Priyanshu, Aman and Vijay, Supriti},
  year={2025},
  howpublished={\url{https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/}},
  note={Interactive analysis tool for expert activation patterns in MoE architectures}
}

References & Resources

Downloads last month
4
Safetensors
Model size
9.56B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Dataset used to train AmanPriyanshu/gpt-oss-9.6b-specialized-math-pruned-moe-only-13-experts

Collection including AmanPriyanshu/gpt-oss-9.6b-specialized-math-pruned-moe-only-13-experts