DataLearner logo
GP

GPT OSS 20B

Reasoning modelGPT OSSGPT OSS 20B

GPT Opensources 20B

Release date: 2025-08-06Updated: 2026-06-15Views: 1,666
Live demoGitHubHugging FaceCompare
Parameters
21B
Context length
128K
Multilingual
Supported
Reasoning ability
3/5

GPT Opensources 20B is an AI model published by OpenAI, released on 2025-08-06, for Reasoning model, with 21B parameters, and 128K context length, requiring about 42GB storage, with a 2516.00 score on CodeForces.

Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology

GPT OSS 20B

Model basics

Reasoning traces
Supported
Thinking modes
Standard Mode (Default)Thinking Mode
Context length
128K tokens
Max output length
4K tokens
Model type
Reasoning model
Modality (in / out)
Text → Text
Release date
2025-08-06
Model file size
42GB
MoE architecture
Yes
Total params / Active params
21B / 3.6B
Knowledge cutoff
No data
GPT OSS 20B

Open source & experience

Code license
Weights license
Apache 2.0- Commercial use permitted
GitHub repo
N/A
Live demo
N/A
GPT OSS 20B

Official resources

Paper
DataLearnerAI blog
N/A
GPT OSS 20B

API details

API speed
2/5
💡Default unit: $/1M tokens. If vendors use other units, follow their published pricing.
Standard
TypeConditionInputOutput
Text-$0.070/ 1M tokens$0.300/ 1M tokens
Batch
TypeConditionInputOutput
Text-$0.035/ 1M tokens$0.150/ 1M tokens
Cache PricingPrompt Cache
TypeTTLWriteRead
Text-$0.070/ 1M tokens$0.040/ 1M tokens
Text-$0.070/ 1M tokens
Cache = write
Text-$0.040/ 1M tokens
Cache = hit

“—” means the modality is not billed in that direction, or the vendor has not published a price for it.

GPT OSS 20B

Benchmark Results

GPT OSS 20B currently shows benchmark results led by AIME 2024 (3 / 62, score 96), AIME2025 (14 / 107, score 98.70), MMLU (41 / 124, score 85.30). This page also consolidates core specs, context limits, and API pricing so you can evaluate the model from benchmark results and deployment constraints together.

Thinking
Tool usage

General Knowledge

4 evaluations
Benchmark / mode
Score
Rank/total
MMLU
Thinking Mode
85.30
41 / 124
MMLU Pro
Thinking Mode
74
87 / 134
HLE
Thinking Mode
10.90
167 / 197
HLE
Thinking ModeTools
17.30
153 / 197

General Evaluation

4 evaluations
Benchmark / mode
Score
Rank/total
53.16
242 / 274
60.80
226 / 274
GPQA Diamond
Thinking Mode
71.50
187 / 274
45.96
253 / 274

Coding and Software Engineer

3 evaluations
Benchmark / mode
Score
Rank/total
CodeForces
Thinking Mode
2230
13 / 21
CodeForces
Thinking ModeTools
2516
10 / 21
SWE-bench Verified
Thinking Mode
34
110 / 116

Math and Reasoning

3 evaluations
Benchmark / mode
Score
Rank/total
AIME2025
Thinking Mode
79
60 / 107
AIME2025
Thinking ModeTools
98.70
14 / 107
AIME 2024
Thinking ModeTools
96
3 / 62

Writing and Creative Capabilities

1 evaluations
Benchmark / mode
Score
Rank/total
Creative Writing
Standard Mode
663.70
104 / 106

Agent Level Benchmark

1 evaluations
Benchmark / mode
Score
Rank/total
τ²-Bench
Thinking ModeTools
47.70
37 / 44

Instruction Following

1 evaluations
Benchmark / mode
Score
Rank/total
65.10
25 / 36

AI Agent - Information Search

1 evaluations
Benchmark / mode
Score
Rank/total
BrowseComp
Thinking ModeTools
28.30
54 / 57

Claw-style Agent Evaluation

1 evaluations
Benchmark / mode
Score
Rank/total
Pinch Bench
Thinking ModeTools
66
35 / 38
GPT OSS 20B

Publisher

GPT Opensources 20B

Model Overview

GPT Opensources 20B is an AI model published by OpenAI, released on 2025-08-06, for Reasoning model, with 21B parameters, and 128K context length, requiring about 42GB storage, with a 2516.00 score on CodeForces.

DataLearner on WeChat

Follow DataLearner on WeChat for AI model updates and research notes.

DataLearner WeChat QR code