Project: https://amanpriyanshu.github.io/GPT-OSS-MoE-ExpertFingerprinting/
This is a pruned variant of OpenAI's GPT-OSS-20B model, reduced to 6 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 all 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.
| Metric | Value |
|---|---|
| Base Model | openai/gpt-oss-20b |
| Architecture | Mixture-of-Experts Transformer |
| Total Parameters | ~5.4B (pruned from 21B) |
| Original Experts per Layer | 32 |
| Pruned Experts per Layer | 6 |
| 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 | All |
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:
Note: Performance may vary depending on how well the pruned experts match your specific use case.
This general-purpose model maintains broad capabilities across all domains while significantly reducing computational requirements. It preserves the essential routing patterns discovered across our comprehensive analysis of diverse evaluation benchmarks including GPQA, MMLU, SORRY-Bench, and Tulu3 datasets.
The expert selection process utilized our comprehensive analysis of router activation patterns across multiple evaluation benchmarks:
By identifying experts that consistently activated for all tasks, we created this specialized model that maintains domain expertise while significantly reducing computational requirements from 32 to 6 experts per layer.
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.
Our approach involves:
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.
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
# Load the specialized model on CPU
model = AutoModelForCausalLM.from_pretrained(
"AmanPriyanshu/gpt-oss-5.4b-specialized-all-pruned-moe-only-6-experts",
torch_dtype=torch.bfloat16,
device_map="cpu",
trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("AmanPriyanshu/gpt-oss-5.4b-specialized-all-pruned-moe-only-6-experts")
# Generate with the model
messages = [
{"role": "user", "content": "What is artificial intelligence and how does it work?"}
]
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)
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-5.4b-specialized-all-pruned-moe-only-6-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-5.4b-specialized-all-pruned-moe-only-6-experts")
# Generate with the model
messages = [
{"role": "user", "content": "What is artificial intelligence and how does it work?"}
]
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)
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
The following examples demonstrate the model's performance on all tasks:
Question: What is artificial intelligence and how does it work?
Response:
<|channel|>analysis<|message|>Need to explain AI concept, how it works. Provide explanation.<|end|><|start|>assistant<|channel|>final<|message|>### What is Artificial Intelligence?
Artificial Intelligence (AI) is a **machine‑based** system that can **learn, reason, and act** on a **human‑like** level. In short, it’s a “computer‑brain” that can understand, predict, and solve problems in a way that a human could do.
---
## 1. What does it work?
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
Question: Explain Einstein's theory of relativity in simple terms.
Response:
<|channel|>analysis<|message|>We need to explain Einstein's theory of relativity in simple terms. Provide a concise explanation.<|end|><|start|>assistant<|channel|>final<|message|>**Einstein’s theory of relativity is a way to make sure that two‑particles (or two‑objects) can move in space, but not in time‑space.**
---
### 1. What the problem is?
- **Two‑objects** (e.g., a particle and a particle) can be in the same space (the same physical dimension).
- **Time‑space** (the time in the same space) is also the same dimension.
- If you can move a particle from a space that is in the same dimension, you must keep the same time‑space.
- In classical physics, if you move a particle from a space, you must change the time‑space, which is not possible.
### 2. The solution:
- **The idea is that the time‑space is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is a “time‑space” that is
Question: How does democracy function in modern governments?
Response:
<|channel|>analysis<|message|>User asks: "How does democracy function in modern governments?" They want explanation of how democracy functions in modern governments. Provide answer.<|end|><|start|>assistant<|channel|>final<|message|>### How does democracy function in modern governments?
**1. Representation of the people**
- **Demlegion**: The people elect representatives (or vote directly) that they are accountable.
- **Checks**: The representatives can be removed or replaced if they fail to act in the public interest.
**2. Decision‑making**
- **Legislation**: The laws are drafted, debated, and voted.
- **Policy**: The public can propose, debate, and vote on policy proposals.
**3. Checks and balances**
- **Parim**: The legislature, the judiciary, and the executive branch are all‑checked.
- **Checks**: The public can call the media, the press, and the courts.
**4. Transparency and accountability**
- **Public**: The public can read the documents, the public can read the public.
- **Transparency**: The public can read the public.
**5. Citizen participation**
- **Public**: The public can read the public.
- **Citizen**: The public can read the public.
**6. Checks and balances**
- **Parim**: The legislature, the judiciary, and the executive branch are all‑checked.
- **Checks**: The public can call the media, the press, and the courts.
**7. Citizen participation**
- **Public**: The public can read the public.
- **Citizen**: The public can read the public.
**8. Checks and balances**
- **Parim**: The legislature, the judiciary, and the executive branch are all‑checked.
- **Checks**: The public can call the media, the press, and the courts.
**9. Citizen participation**
- **Public**: The public can read the public.
- **Citizen**: The public can read the public.
**10. Checks and balances**
- **Parim**: The legislature, the judiciary, and the executive branch are all‑checked.
- **Checks**: The public can call the media, the press, and the courts.
**11. Citizen participation**
- **Public**: The public can read the public.
- **Citizen**: The public can read
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}
}