Better than 46% of all models
Context
131K tokens (~66 books)
Input $/1M
$0.03
Output $/1M
$0.14
Type
text
License
Open Source
Benchmarks
6 tested
Data updated today
About
gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license. It uses a Mixture-of-Experts (MoE) architecture with 3.6B active parameters per forward pass, optimized for...
Tested on 6 benchmarks with 67.4% average. Top scores: Chatbot Arena Elo — Overall (1317.7%), HELM — MMLU-Pro (74.0%), HELM — WildBench (73.7%).
Capabilities
reasoning
73.7
#18 globally
math
56.5
#58 globally
knowledge
66.7
#26 globally
language
73.2
#64 globally
Benchmark Scores
Compare AllTested on 6 benchmarks · Ranked across 5 categories
Score Distribution (all 233 models)
0255075100
▲ You are here
reasoningCompare reasoning →
HELM — WildBench
73.7—Stanford HELM WildBench evaluation. Tests reasoning on challenging real-world tasks.
mathCompare math →
HELM — Omni-MATH
56.5—Stanford HELM evaluation of mathematical reasoning across diverse problem types.
knowledgeCompare knowledge →
HELM — MMLU-Pro
74.0—Stanford HELM evaluation of MMLU-Pro. Tests broad knowledge with increased difficulty.
Excellent (85+) Good (70-85) Average (50-70) Below (<50)
Links
Research
Documentation
Community
Source Code
BenchGecko API
gpt-oss-20b
Specifications
- Typetext
- Context131K tokens (~66 books)
- ReleasedAug 2025
- LicenseOpen Source
- StatusActive
- Cost / Message~$0.000
Available On
Share & Export
Frequently Asked Questions
gpt-oss-20b is an open-source text AI model by OpenAI, released in August 2025. It has an average benchmark score of 44.4. Context window: 131K tokens.