RankingEffort benchmarks

gpt-oss-120b effort benchmarks

Compare measured Low, Medium, High and other settings. Every number keeps its source.

Model

gpt-oss-120b

45 measured results · 3 reported effort labels · 28 benchmark/harness combinations. Model evidence and sources →

Capability score evidence

gpt-oss-120b · High: 4.5 overall comparison score. Meets the overall evidence requirements.

Which sources carry the weight?

Artificial Analysis · 48.1% direct comparison weight · independent
SWE-rebench · 36.0% direct comparison weight · independent
Epoch AI · 15.9% direct comparison weight · independent

These shares describe the comparisons entering the fit, not fractions of the final score.

How sensitive is this score?

Remove one evaluator or family and refit the graph against the same overall panel. Values use the matching dataset audited 2026-09-12. These are sensitivity checks, not confidence intervals; ranking eligibility may also change.

Removed evidence	Recomputed score	Change
Artificial Analysis	1.6	-2.9
ARC Prize	5.3	+0.9
Datacurve	3.8	-0.7
LiveCodeBench	4.4	-0.0
Scale	5.0	+0.5
Terminal-Bench	4.4	-0.0
Cognition	4.7	+0.2
LiveBench	4.5	+0.1
SWE-rebench	6.3	+1.9
Epoch AI	5.7	+1.3
Vals AI	4.3	-0.1
Provider reports · Anthropic	4.5	+0.0
Provider reports · Moonshot AI	4.5	-0.0
Provider reports · Meta	4.4	-0.0
Provider reports · Mistral	4.5	+0.0
All provider reports	4.4	-0.0

Every benchmark family

Removed evidence	Recomputed score	Change
swe-rebench	6.3	+1.9
epoch-game-puzzles	5.4	+1.0
arc-agi	5.3	+0.8
deepswe	3.8	-0.7
mmmu-pro	4.8	+0.3
frontiercode	4.7	+0.2
ifbench	4.2	-0.2
simpleqa	4.6	+0.2
mmlu-pro	4.6	+0.2
aime	4.3	-0.2
gpqa	4.3	-0.2
briefcase	4.6	+0.2
livecodebench	4.3	-0.1
scicode	4.6	+0.1
automationbench	4.6	+0.1
omniscience	4.3	-0.1
hle	4.6	+0.1
critpt	4.4	-0.1
tau-bench	4.4	-0.1
itbench	4.5	+0.1
apex-agents	4.5	+0.1
livebench-instructions	4.5	+0.1
multi-swe-bench	4.4	-0.1
enterprise-ops	4.5	+0.1
livebench-language	4.5	+0.1
swe-atlas-qna	4.4	-0.0
livebench-coding	4.4	-0.0
arc-agi-3	4.4	-0.0
harvey	4.5	+0.0
livebench-reasoning	4.4	-0.0
vibe-code	4.4	-0.0
frontiermath	4.4	-0.0
vals-finance-agent	4.4	-0.0
terminal-bench	4.5	+0.0
vals-legal-research	4.4	-0.0
gmmlu	4.4	-0.0
swe-atlas-test-writing	4.4	-0.0
vals-code-migration	4.4	-0.0
gdp-pdf	4.5	+0.0
math500	4.5	+0.0
analyst-agent	4.5	+0.0
livebench-data	4.5	+0.0
vals-excel-modeling	4.4	-0.0
enigma-eval	4.5	+0.0
livebench-math	4.5	+0.0
swe-atlas-refactoring	4.4	-0.0
aa-lcr	4.5	-0.0
gdpval	4.5	-0.0
swe-bench-pro	4.5	+0.0

What evidence is missing from the fit?

Across all collected settings for this model: 48 observations, 43 contributing, 0 matched without graph weight and 5 excluded. Counts do not establish rank eligibility.

2 · Benchmark family or source protocol has not been reviewed
3 · No reviewed effort for this catalog observation; any separately collected effort measurement is counted separately

Inspect every observation and exclusion →

Capability profile

One configuration, seven capabilities

Comparison score · 0–100

Agpt-oss-120b · High4 of 7 capabilities supported

Gaps are unknown, not zero. Hollow points are preliminary.

Agentic
A6.2Supported
Hard reasoning
A11.6Supported
Coding
A5.7Supported
Human pref
Unknown
Knowledge
A12.2Supported
Multimodal
Unknown
Long context
A8.9Preliminary

Each profile belongs to the named configuration. Capability scores use a shared panel for all axes, distinct from the overall ranking panel; they are not accuracy percentages or direct head-to-head wins.

Compare this configuration →Explore the model evidence profile →

These are published benchmark scores. Supported effort estimates also appear on the Capability leaderboard. Results from different harnesses occupy different rows. Missing cells stay unknown; no scores are copied between effort levels.

Benchmark / source	Low	Medium	High
AA-Briefcase EloArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details AA-Briefcase is an agentic knowledge work benchmark developed by Artificial Analysis. AA-Briefcase Elo is a combined metric that aggregates rubric pass rate, analytical quality Elo and presentation Elo · Higher is better · Evaluation results measured independently by Artificial Analysis Unit: elo. Source collected 2026-09-12.	Not reported	Not reported	0 Elo Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
AIME 2025Artificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	66.67% Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	93.44% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
APEX-Agents-AAArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	Not reported	3.1% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
AA-LCR v1.1Artificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	46% Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	52% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
AutomationBench-AAArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Share of task objectives completed with no guardrail violations · Higher is better · Benchmark developed by Zapier · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	Not reported	0.2% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
CritPtArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Benchmark developed by Argonne and UIUC, with contributions from 60+ researchers globally · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	0% Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	1.14% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
EnterpriseOps-Gym-AAArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Overall task success rate across all eight enterprise domains (oracle tool mode) · Benchmark developed by ServiceNow Research · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	Not reported	25.54% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
GDP.pdf: All-passArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Share of attempts where every atomic criterion passed · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	Not reported	4% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
GDPval-AA v2 LeaderboardArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Elo rating for performance on real-world work tasks · Anchored to a human baseline of 1,000 · Higher is better · Evaluation results measured independently by Artificial Analysis Unit: elo. Source collected 2026-09-12.	364.46 Elo Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	744.74 Elo Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
Global-MMLU-LiteArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	80.12% Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	82.78% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
GPQA DiamondArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	67.17% Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	78.18% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
Harvey LAB-AA: Criterion Pass RateArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Share of atomic pass/fail rubric criteria the deliverables satisfy, graded by a single LLM judge · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	Not reported	13.89% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
Humanity's Last ExamArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	5.89% Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	19.6% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
IFBenchArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	58.3% Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	68.98% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
ITBench-AAArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Combined public and private Site Reliability Engineering (SRE) tasks · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	Not reported	5.65% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
LiveCodeBenchArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	70.69% Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	87.83% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
MLCR-AAArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Share of tasks judged accurate, complete, and concise · Expert + compound (hard) sets · Benchmark developed by Wisedocs · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	Not reported	1.11% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
MMLU-ProArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	77.48% Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	80.79% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
AA-Omniscience IndexArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details AA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct. · Evaluation results measured independently by Artificial Analysis Unit: index. Source collected 2026-09-12.	-53.5 Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	-49.25 Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
AA-Omniscience AccuracyArtificial Analysis · 6000 questions; 1 attempt; correct-answer share; GPT-5.6 Luna medium grader Benchmark details AA-Omniscience Accuracy (higher is better) measures the proportion of correctly answered questions out of all questions, regardless of whether the model chooses to answer · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	19.8% Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	21.78% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
SciCodeArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	Not reported	34.03% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
𝜏²-Bench TelecomArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	45.03% Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	65.79% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
𝜏³-BankingArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Benchmark developed by Sierra Research · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	2.89% Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	12.78% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
Terminal-Bench HardArtificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	5.3% Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	23.48% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
Terminal-Bench v2.1Artificial Analysis · Artificial Analysis published evaluation; see source for benchmark-specific settings Benchmark details Benchmark developed by the Laude Institute and open-source contributors · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	13.86% Reported settings gpt-oss-120b (low); Model source: https://artificialanalysis.ai/models/gpt-oss-120b-low	Not reported	26.22% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
Terminal-Bench v4.0Artificial Analysis · mini-SWE-agent v2.4.6; 66 tasks; pass@1 averaged over 3 repeats; 500 steps; task timeout up to 8 hours Benchmark details Benchmark developed by the Laude Institute, Stanford, and open-source contributors · Independently benchmarked by Artificial Analysis · Evaluation results measured independently by Artificial Analysis Unit: percent. Source collected 2026-09-12.	Not reported	Not reported	0% Reported settings gpt-oss-120b (high); Model source: https://artificialanalysis.ai/models/gpt-oss-120b
SWE-rebench · 2025-12 · toolsSWE-rebench · SWE-rebench standardized ReAct tools Benchmark details Average resolved rate across repeated runs, not best-of-N success. Disjoint complete calendar months within each model’s tested range; standardized harness and all-language task set. Unit: percent. Source collected 2026-09-12.	Not reported	Not reported	36.96% Reported settings SWE-rebench standardized ReAct tools; swe-rebench-2025-12-tools; gpt-oss-120b-high; checkpoint=gpt-oss-120b
Epoch · Mystery Game Puzzles · v1.0.4 · Tool submissionEpoch AI · Epoch AI Inspect; task 1.0.4; scorer agent_submitted_next_move_scorer Benchmark details Epoch AI internal evaluation. Mean accuracy under the published setting; exact task version and scorer stay separate. Related tasks share a family budget. Data by Epoch AI, CC BY. Unit: percent. Source collected 2026-09-12.	Not reported	2% Reported settings Epoch AI Inspect; task 1.0.4; scorer agent_submitted_next_move_scorer; model=gpt-oss-120b_medium; run=58MQQiJ7vvgvP9e8LcG6eV; mean accuracy; https://epoch.ai/data/benchmarks.csv	0% Reported settings Epoch AI Inspect; task 1.0.4; scorer agent_submitted_next_move_scorer; model=gpt-oss-120b_high; run=GbQrTQyihBUsb9sxbqT8pT; mean accuracy; https://epoch.ai/data/benchmarks.csv

What these comparisons mean

Each source reports its own effort labels. Matching labels do not establish equal compute budgets or identical fallback behavior. Open Reported settings to inspect the exact model name, agent and task coverage. Multiple results at one level remain visible rather than selecting the best. Source-reported None or Non-reasoning labels are not treated as evidence that an API supports those settings.

These measurements use their own dated source collection. The leaderboard fits identified configurations jointly; it contains no pooled model ratings. Explicit thinking budgets are separate configurations. Read the methodology.

Download all effort benchmark results · 7,424 results across 240 models · Collected 2026-09-12

gpt-oss-120b effort benchmarks

gpt-oss-120b

Capability score evidence

One configuration, seven capabilities

Agentic

Hard reasoning

Coding

Human pref

Knowledge

Multimodal

Long context