GPT-OSS 20B
Active Parameters
21B
Context Length
128K
Modality
Text
Architecture
Mixture of Experts (MoE)
License
Apache 2.0
Release Date
5 Aug 2025
Knowledge Cutoff
Jun 2024
API Pricing (per 1M)
Input: $0.06 · Output: $0.19
System Requirements
VRAM requirements for different quantization methods and context sizes
1,024 tokens
Consumer
2x RTX 4090
24GB VRAM
Datacenter
1x NVIDIA A100
80GB VRAM
Apple Silicon
1x Apple M3 Max
128GB VRAM
128,000 tokens
Consumer
3x RTX 4090
24GB VRAM
Datacenter
1x NVIDIA A100
80GB VRAM
Apple Silicon
1x Apple M3 Max
128GB VRAM
Architecture Diagram
Evaluation Benchmarks
Rank
#115
| Benchmark | Score | Rank |
|---|---|---|
Graduate-Level QA | high 0.742 Standard 0.715 | 77 79 |
Agentic Index | high 0.01 | 111 |
General Text | 1317 | 128 |
Coding Index | high 0.21 | 132 |
Intelligence Index | high 0.09 low 0.10 | 187 178 |
Summarization Archived | 0.863 | 6 |
General Knowledge Reference | 0.853 | 11 |
Rankings
Overall Rank
#115
Coding Rank
#118
About GPT-OSS 20B
GPT-OSS 20B is an open-weight Mixture-of-Experts language model by OpenAI designed for efficient local reasoning on consumer hardware. It activates 3.6B parameters per token and features native tool use across a 128K context window.
Technical Specifications
Attention
Attention Structure
Multi-Head Attention
Attention Heads
64
Key-Value Heads
8
Attention Head Dimension
64
Position Embedding
Absolute Position Embedding
RoPE Theta
150,000
Sliding Window Attention
Yes
Sliding Window Size
128
Sliding Window Ratio
-
Linear Attention
-
Linear Attention Ratio
-
Normalization
RMS Normalization
Activation Function
SwigLU
Dimensions
Hidden Dimension Size
2,880
Number of Layers
24
FFN Intermediate Size (Dense)
2,880
Multi-Token Prediction Heads
-
Tokenizer
Vocabulary Size
201,088
Mixture of Experts
Total Expert Parameters
3.6B
Number of Experts
32
Active Experts
4
Shared Experts
-
FFN Intermediate Size (per Expert)
-
Dense Layers Before MoE
-