성능 벤치마크
여러 평가 도메인에서 최첨단 결과
추론, 코딩, 수학 및 전문 분야에서 GPT-OSS 모델의 상세한 성능 지표와 비교를 살펴보세요.
성능 개요
GPT-OSS 모델은 다양한 벤치마크에서 뛰어난 성능을 보여주며, 오픈 웨이트 언어 모델의 새로운 기준을 설정합니다.
85%
MMLU 점수
GPT-OSS 120B
2622
Codeforces 레이팅
경쟁 프로그래밍
131k
컨텍스트 길이
확장된 컨텍스트 윈도우
Apache 2.0
라이선스
완전 오픈소스
performance.benchmarks.title
performance.benchmarks.description
performance.benchmarks.categories.0.title
performance.benchmarks.categories.0.description
| Benchmark | Description | GPT-OSS 120B | GPT-OSS 20B | Baseline |
|---|---|---|---|---|
| performance.benchmarks.categories.0.tests.0.name | performance.benchmarks.categories.0.tests.0.description | performance.benchmarks.categories.0.tests.0.gpt_oss_120b | performance.benchmarks.categories.0.tests.0.gpt_oss_20b | performance.benchmarks.categories.0.tests.0.baseline |
| performance.benchmarks.categories.0.tests.1.name | performance.benchmarks.categories.0.tests.1.description | performance.benchmarks.categories.0.tests.1.gpt_oss_120b | performance.benchmarks.categories.0.tests.1.gpt_oss_20b | performance.benchmarks.categories.0.tests.1.baseline |
| performance.benchmarks.categories.0.tests.2.name | performance.benchmarks.categories.0.tests.2.description | performance.benchmarks.categories.0.tests.2.gpt_oss_120b | performance.benchmarks.categories.0.tests.2.gpt_oss_20b | performance.benchmarks.categories.0.tests.2.baseline |
performance.benchmarks.categories.1.title
performance.benchmarks.categories.1.description
| Benchmark | Description | GPT-OSS 120B | GPT-OSS 20B | Baseline |
|---|---|---|---|---|
| performance.benchmarks.categories.1.tests.0.name | performance.benchmarks.categories.1.tests.0.description | performance.benchmarks.categories.1.tests.0.gpt_oss_120b | performance.benchmarks.categories.1.tests.0.gpt_oss_20b | performance.benchmarks.categories.1.tests.0.baseline |
| performance.benchmarks.categories.1.tests.1.name | performance.benchmarks.categories.1.tests.1.description | performance.benchmarks.categories.1.tests.1.gpt_oss_120b | performance.benchmarks.categories.1.tests.1.gpt_oss_20b | performance.benchmarks.categories.1.tests.1.baseline |
| performance.benchmarks.categories.1.tests.2.name | performance.benchmarks.categories.1.tests.2.description | performance.benchmarks.categories.1.tests.2.gpt_oss_120b | performance.benchmarks.categories.1.tests.2.gpt_oss_20b | performance.benchmarks.categories.1.tests.2.baseline |
performance.benchmarks.categories.2.title
performance.benchmarks.categories.2.description
| Benchmark | Description | GPT-OSS 120B | GPT-OSS 20B | Baseline |
|---|---|---|---|---|
| performance.benchmarks.categories.2.tests.0.name | performance.benchmarks.categories.2.tests.0.description | performance.benchmarks.categories.2.tests.0.gpt_oss_120b | performance.benchmarks.categories.2.tests.0.gpt_oss_20b | performance.benchmarks.categories.2.tests.0.baseline |
| performance.benchmarks.categories.2.tests.1.name | performance.benchmarks.categories.2.tests.1.description | performance.benchmarks.categories.2.tests.1.gpt_oss_120b | performance.benchmarks.categories.2.tests.1.gpt_oss_20b | performance.benchmarks.categories.2.tests.1.baseline |
| performance.benchmarks.categories.2.tests.2.name | performance.benchmarks.categories.2.tests.2.description | performance.benchmarks.categories.2.tests.2.gpt_oss_120b | performance.benchmarks.categories.2.tests.2.gpt_oss_20b | performance.benchmarks.categories.2.tests.2.baseline |
performance.benchmarks.categories.3.title
performance.benchmarks.categories.3.description
| Benchmark | Description | GPT-OSS 120B | GPT-OSS 20B | Baseline |
|---|---|---|---|---|
| performance.benchmarks.categories.3.tests.0.name | performance.benchmarks.categories.3.tests.0.description | performance.benchmarks.categories.3.tests.0.gpt_oss_120b | performance.benchmarks.categories.3.tests.0.gpt_oss_20b | performance.benchmarks.categories.3.tests.0.baseline |
| performance.benchmarks.categories.3.tests.1.name | performance.benchmarks.categories.3.tests.1.description | performance.benchmarks.categories.3.tests.1.gpt_oss_120b | performance.benchmarks.categories.3.tests.1.gpt_oss_20b | performance.benchmarks.categories.3.tests.1.baseline |
| performance.benchmarks.categories.3.tests.2.name | performance.benchmarks.categories.3.tests.2.description | performance.benchmarks.categories.3.tests.2.gpt_oss_120b | performance.benchmarks.categories.3.tests.2.gpt_oss_20b | performance.benchmarks.categories.3.tests.2.baseline |
performance.efficiency.title
performance.efficiency.description
performance.efficiency.metrics.0.title
performance.efficiency.metrics.0.description
GPT-OSS 120Bperformance.efficiency.metrics.0.gpt_oss_120b
GPT-OSS 20Bperformance.efficiency.metrics.0.gpt_oss_20b
Hardware: performance.efficiency.metrics.0.hardware
performance.efficiency.metrics.1.title
performance.efficiency.metrics.1.description
GPT-OSS 120Bperformance.efficiency.metrics.1.gpt_oss_120b
GPT-OSS 20Bperformance.efficiency.metrics.1.gpt_oss_20b
Hardware: performance.efficiency.metrics.1.hardware
performance.efficiency.metrics.2.title
performance.efficiency.metrics.2.description
GPT-OSS 120Bperformance.efficiency.metrics.2.gpt_oss_120b
GPT-OSS 20Bperformance.efficiency.metrics.2.gpt_oss_20b
Hardware: performance.efficiency.metrics.2.hardware
performance.methodology.title
performance.methodology.description
1
performance.methodology.principles.0.title
performance.methodology.principles.0.description
2
performance.methodology.principles.1.title
performance.methodology.principles.1.description
3
performance.methodology.principles.2.title
performance.methodology.principles.2.description
4
performance.methodology.principles.3.title
performance.methodology.principles.3.description
performance.comparison.title
performance.comparison.description
performance.comparison.note
performance.cta.title
performance.cta.description