OpenAIproprietary
Released: 2026-06-01

OpenAI: gpt-oss-20b

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...

Intelligence Index
77.0/ 100
Calibrated multi-domain composite
Throughput Speed
130tok/sec
Output streaming throughput
Time To First Token
280ms
Initial chunk server latency
LiveBench Contam-Free
67.8%
September 2026 suite score

LiveBench Multi-Domain Evaluation Breakdown

Monthly refreshed contamination-free test sets
Logical Reasoning70.8%
Coding & Repo Repair69.3%
Mathematics (AIME)68.5%
Data Analysis & Tables65.5%
Language & Comprehension67.0%
Instruction Following67.8%

Sourced Benchmark Evaluations (61)

Verified performance across authoritative benchmarks with provenance tracking.
BenchmarkStatusScoreTrustDateSource TypeProvenance
ARC-AGI-3Active48.6%
93
2026-07-06independentSource ↗
ARC-AGI-2Nearing Saturation58.3%
75
2026-06-29independentSource ↗
FrontierMathActive20%
98
2026-06-26independentSource ↗
PutnamBenchActive37.2%
90
2026-06-25independentSource ↗
SWE-LancerDeprecated49.7%
20
2026-06-23independentSource ↗
Omni-MATHDeprecated53%
10
2026-06-23independentSource ↗
SoohakActive54.4%
98
2026-06-21independentSource ↗
FrontierCodeActive42.4%
98
2026-06-21independentSource ↗
Humanity's Last ExamActive32.9%
93
2026-06-19independentSource ↗
StructFlowBenchNearing Saturation57.6%
75
2026-06-19independentSource ↗
AIMENearing Saturation68%
75
2026-06-19independentSource ↗
OlympiadBenchActive58.7%
80
2026-06-18independentSource ↗
MathArenaActive57.3%
99
2026-06-17independentSource ↗
FireBenchActive57%
95
2026-06-17independentSource ↗
Terminal-BenchActive51.2%
80
2026-06-17independentSource ↗
OpenAI-MRCRActive64.7%
80
2026-06-16independentSource ↗
SimpleQAActive40%
90
2026-06-16independentSource ↗
CodeContestsDeprecated47.5%
10
2026-06-16independentSource ↗
ComplexBenchNearing Saturation55.6%
80
2026-06-15independentSource ↗
SWE-bench ProActive45.3%
90
2026-06-15independentSource ↗
LongBench-ProActive63.1%
80
2026-06-15independentSource ↗
ZeroSCROLLSNearing Saturation61.6%
78
2026-06-14independentSource ↗
MultiPL-EDeprecated67.3%
10
2026-06-14independentSource ↗
LiveCodeBenchSaturated59.9%
39
2026-06-13independentSource ↗
AgentIFNearing Saturation46.1%
80
2026-06-13independentSource ↗
MATHSaturated70.2%
25
2026-06-13independentSource ↗
InfiniteBenchNearing Saturation65.5%
65
2026-06-13independentSource ↗
MercuryDeprecated64.3%
10
2026-06-12independentSource ↗
SCROLLSNearing Saturation61.6%
65
2026-06-12independentSource ↗
MMLU-ProNearing Saturation70.8%
75
2026-06-12independentSource ↗
RULERNearing Saturation70.8%
84
2026-06-11independentSource ↗
TabMWPSaturated65.9%
25
2026-06-11independentSource ↗
CyberSecEvalDeprecated65.8%
10
2026-06-11independentSource ↗
SWE-bench VerifiedSaturated57%
25
2026-06-10independentSource ↗
MMLU-ReduxSaturated75.5%
25
2026-06-10independentSource ↗
QuALITYNearing Saturation67.8%
78
2026-06-10independentSource ↗
LongBenchNearing Saturation66.2%
65
2026-06-09independentSource ↗
SWE-benchDeprecated49.7%
10
2026-06-09independentSource ↗
MGSMNearing Saturation68%
65
2026-06-09independentSource ↗
AGIEvalSaturated64.2%
25
2026-06-08independentSource ↗
QASPERNearing Saturation63.9%
65
2026-06-08independentSource ↗
TruthfulQASaturated67.8%
25
2026-06-08independentSource ↗
GSM-HardDeprecated68.7%
10
2026-06-07independentSource ↗
SuperGLUESaturated76.2%
25
2026-06-07independentSource ↗
CodeXGLUEDeprecated68.7%
23
2026-06-06independentSource ↗
MT-BenchNearing Saturation7%
75
2026-06-06independentSource ↗
BIG-Bench HardSaturated73.2%
25
2026-06-06independentSource ↗
WinoGrandeSaturated77.7%
28
2026-06-05independentSource ↗
MBPP+Nearing Saturation74.6%
65
2026-06-05independentSource ↗
IFEvalDeprecated66.4%
25
2026-06-05independentSource ↗
GPQANearing Saturation67.8%
78
2026-06-05independentSource ↗
Needle-in-a-HaystackSaturated95.4%
44
2026-06-04independentSource ↗
HumanEval+Saturated78.9%
25
2026-06-04independentSource ↗
CommonsenseQASaturated78.4%
28
2026-06-04independentSource ↗
MBPPSaturated77.5%
25
2026-06-03independentSource ↗
GSM8KSaturated77.3%
25
2026-06-03independentSource ↗
GLUESaturated80.7%
28
2026-06-03independentSource ↗
MMLUSaturated78.5%
25
2026-06-02independentSource ↗
HumanEvalSaturated83.3%
25
2026-06-02independentSource ↗
ARC-AGISaturated70.2%
25
2026-06-02independentSource ↗
LiveBench (Continuous)Active67.8%
94
2025-02-01independentSource ↗
OpenAI Compatible API ExecutionAPI Ready
from openai import OpenAI

client = OpenAI()
response = client.chat.completions.create(
    model="openai-gpt-oss-20b",
    messages=[{"role": "user", "content": "Evaluate multi-step logic problem"}],
    temperature=0.2
)
print(response.choices[0].message.content)

Evaluation & Sourcing Notes

Scores listed for OpenAI: gpt-oss-20b represent verified evaluations extracted from official research papers, independent evaluation suites (HELM, LMSYS, OpenCompass, LiveBench), and verified audit reports.

All benchmarks marked as saturated or deprecated reflect historical performance where the benchmark no longer provides active discriminative power.