BenchAtlas

Rankings / OpenAI

GPT 5 Mini (2025-08-07)

proprietary

released 2025-08-07

BenchAtlas Index

as of 2026-07-14
48.9
high effort
rank #225
11 families · 5 categories · high
48.4
base configuration
rank #229
5 families · 4 categories · medium
47.4
high effort
rank #243
14 families · 5 categories · high

Benchmark evidence

125 results
agentic coding
τ²-Benchmedium effort71.0 %independentArtificial Analysis
τ²-Benchhigh effort68.4 %independentArtificial Analysis
τ²-Benchminimal effort31.9 %independentArtificial Analysis
τ²-Bench
subset=banking
high effort14.6 %independentArtificial Analysis
coding
LiveCodeBench v6
implementation=artificial-analysis
high effort83.8 %independentArtificial Analysis
LiveCodeBench v6
implementation=artificial-analysis
medium effort69.2 %independentArtificial Analysis
LiveCodeBench v6
implementation=artificial-analysis
minimal effort54.5 %independentArtificial Analysis
SciCodemedium effort41.0 %independentArtificial Analysis
SciCodehigh effort39.2 %independentArtificial Analysis
SciCodeminimal effort36.9 %independentArtificial Analysis
external indices
AA Coding Indexhigh effort15.6 pointsindependentArtificial Analysis
Intelligence Index v4.1medium effort30.9 pointsindependentArtificial Analysis
Intelligence Index v4.1high effort25.3 pointsindependentArtificial Analysis
Intelligence Index v4.1minimal effort14.3 pointsindependentArtificial Analysis
AA Math Indexhigh effort90.7 pointsindependentArtificial Analysis
AA Math Indexmedium effort85.0 pointsindependentArtificial Analysis
AA Math Indexminimal effort46.7 pointsindependentArtificial Analysis
ECIlow effort145.8 pointsindependentEpoch AI Benchmarking Hub
ECIhigh effort145.8 pointsindependentEpoch AI Benchmarking Hub
ECImedium effort145.8 pointsindependentEpoch AI Benchmarking Hub
ECIminimal effort145.8 pointsindependentEpoch AI Benchmarking Hub
ECI145.8 pointsindependentEpoch AI Benchmarking Hub
factuality
MASK
contamination=Potential contamination warning: This model was evaluated after the public release of MASK, allowing model builder access to the prompts and solutions.
82.6 %
80.3–84.9
independentScale Labs
SimpleQA Verifiedhigh effort21.0 %independentEpoch AI Benchmarking Hub
2025-12-09
human preference
Coding (style control)older version
arena=text · category=coding · style_control=true
high effort1430.5
1422.0–1438.9
communityLMArena Leaderboard Dataset
Industry Software And It Services (style control)older version
arena=text · category=industry_software_and_it_services · style_control=true
high effort1415.5
1408.6–1422.5
communityLMArena Leaderboard Dataset
Hard Prompts English (style control)older version
arena=text · category=hard_prompts_english · style_control=true
high effort1410.0
1402.1–1417.9
communityLMArena Leaderboard Dataset
Industry Mathematical (style control)older version
arena=text · category=industry_mathematical · style_control=true
high effort1408.6
1392.1–1425.2
communityLMArena Leaderboard Dataset
Chinese (style control)older version
arena=text · category=chinese · style_control=true
high effort1407.8
1390.8–1424.7
communityLMArena Leaderboard Dataset
Math (style control)older version
arena=text · category=math · style_control=true
high effort1405.2
1389.8–1420.6
communityLMArena Leaderboard Dataset
Expert (style control)older version
arena=text · category=expert · style_control=true
high effort1403.5
1385.4–1421.6
communityLMArena Leaderboard Dataset
Hard Prompts (style control)older version
arena=text · category=hard_prompts · style_control=true
high effort1402.8
1396.6–1408.9
communityLMArena Leaderboard Dataset
Industry Medicine And Healthcare (style control)older version
arena=text · category=industry_medicine_and_healthcare · style_control=true
high effort1402.1
1385.8–1418.4
communityLMArena Leaderboard Dataset
Polish (style control)older version
arena=text · category=polish · style_control=true
high effort1397.7
1381.2–1414.3
communityLMArena Leaderboard Dataset
English (style control)older version
arena=text · category=english · style_control=true
high effort1397.6
1391.6–1403.7
communityLMArena Leaderboard Dataset
Industry Life And Physical And Social Science (style control)older version
arena=text · category=industry_life_and_physical_and_social_science · style_control=true
high effort1396.2
1386.6–1405.8
communityLMArena Leaderboard Dataset
Industry Legal And Government (style control)older version
arena=text · category=industry_legal_and_government · style_control=true
high effort1394.5
1379.7–1409.3
communityLMArena Leaderboard Dataset
Overall (style control)
arena=text · category=overall · style_control=true
high effort1390.0
1385.4–1394.6
communityLMArena Leaderboard Dataset
German (style control)older version
arena=text · category=german · style_control=true
high effort1388.8
1363.3–1414.2
communityLMArena Leaderboard Dataset
Non English (style control)older version
arena=text · category=non_english · style_control=true
high effort1379.4
1373.5–1385.3
communityLMArena Leaderboard Dataset
Industry Business And Management And Financial Operations (style control)older version
arena=text · category=industry_business_and_management_and_financial_operations · style_control=true
high effort1378.9
1369.8–1388.0
communityLMArena Leaderboard Dataset
Instruction Following (style control)older version
arena=text · category=instruction_following · style_control=true
high effort1374.3
1366.7–1381.9
communityLMArena Leaderboard Dataset
Spanish (style control)older version
arena=text · category=spanish · style_control=true
high effort1373.0
1349.9–1396.2
communityLMArena Leaderboard Dataset
Russian (style control)older version
arena=text · category=russian · style_control=true
high effort1373.0
1357.5–1388.4
communityLMArena Leaderboard Dataset
Multi Turn (style control)older version
arena=text · category=multi_turn · style_control=true
high effort1372.7
1363.3–1382.0
communityLMArena Leaderboard Dataset
Longer Query (style control)older version
arena=text · category=longer_query · style_control=true
high effort1369.5
1361.2–1377.8
communityLMArena Leaderboard Dataset
Exclude Ties (style control)older version
arena=text · category=exclude_ties · style_control=true
high effort1366.2
1359.9–1372.5
communityLMArena Leaderboard Dataset
Industry Writing And Literature And Language (style control)older version
arena=text · category=industry_writing_and_literature_and_language · style_control=true
high effort1352.8
1344.7–1360.8
communityLMArena Leaderboard Dataset
Industry Entertainment And Sports And Media (style control)older version
arena=text · category=industry_entertainment_and_sports_and_media · style_control=true
high effort1349.4
1340.3–1358.4
communityLMArena Leaderboard Dataset
Creative Writing (style control)older version
arena=text · category=creative_writing · style_control=true
high effort1325.4
1315.0–1335.7
communityLMArena Leaderboard Dataset
Korean (style control)older version
arena=text · category=korean · style_control=true
high effort1320.2
1294.6–1345.7
communityLMArena Leaderboard Dataset
knowledge science
GPQA Diamond
implementation=artificial-analysis
high effort82.8 %independentArtificial Analysis
GPQA Diamond
implementation=vals-ai
high effort80.3 %independentVals AI
GPQA Diamond
implementation=artificial-analysis
medium effort80.3 %independentArtificial Analysis
GPQA Diamondhigh effort75.0 %independentEpoch AI Benchmarking Hub
2025-10-30
GPQA Diamondmedium effort71.7 %independentEpoch AI Benchmarking Hub
2025-08-07
GPQA Diamond
implementation=artificial-analysis
minimal effort68.7 %independentArtificial Analysis
Humanity's Last Exam
implementation=artificial-analysis
high effort19.7 %independentArtificial Analysis
Humanity's Last Exam
contamination=Potential contamination warning: This model was evaluated after the public release of HLE, allowing model builder access to the prompts and solutions. · implementation=scale
19.4 %
17.9–21.0
independentScale Labs
Humanity's Last Exam
implementation=artificial-analysis
medium effort14.6 %independentArtificial Analysis
Humanity's Last Exam
implementation=artificial-analysis
minimal effort5.0 %independentArtificial Analysis
MMLU-Pro
implementation=artificial-analysis
high effort83.7 %independentArtificial Analysis
MMLU-Pro
implementation=artificial-analysis
medium effort82.8 %independentArtificial Analysis
MMLU-Pro
implementation=vals-ai
high effort82.2 %independentVals AI
MMLU-Pro
implementation=artificial-analysis
minimal effort77.5 %independentArtificial Analysis
long context instruction
AA-LCRhigh effort68.0 %independentArtificial Analysis
AA-LCRmedium effort66.0 %independentArtificial Analysis
AA-LCRminimal effort35.7 %independentArtificial Analysis
IFBenchhigh effort75.4 %independentArtificial Analysis
IFBenchmedium effort71.2 %independentArtificial Analysis
IFBenchminimal effort45.6 %independentArtificial Analysis
MultiChallengethinking59.0 %
54.9–63.1
independentScale Labs
multimodal
MMMU
implementation=vals-ai
high effort78.9 %independentVals AI
VISTA50.4 %
48.1–52.7
independentScale Labs
professional
CaseLawhigh effort68.5 %independentVals AI
CorpFinhigh effort60.2 %independentVals AI
LegalBenchhigh effort81.8 %independentVals AI
MedQAhigh effort96.1 %independentVals AI
TaxEvalhigh effort75.2 %independentVals AI
reasoning math
AIME
implementation=vals-ai
high effort91.5 %independentVals AI
AIME
year=2025 · implementation=artificial-analysis
high effort90.7 %independentArtificial Analysis
AIME
year=2025 · implementation=artificial-analysis
medium effort85.0 %independentArtificial Analysis
AIME
year=2025 · implementation=artificial-analysis
minimal effort46.7 %independentArtificial Analysis
ARC-AGI-1older version
split=public_eval · model_type=CoT
high effort61.5 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
high effort54.3 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=public_eval · model_type=CoT
medium effort46.3 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
medium effort37.3 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
low effort26.3 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=public_eval · model_type=CoT
low effort24.4 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=public_eval · model_type=CoT
minimal effort7.1 %independentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
high effort5.8 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
minimal effort5.3 %independentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
high effort4.4 %independentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
medium effort4.0 %independentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
minimal effort1.7 %independentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
low effort0.8 %independentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
medium effort0.6 %independentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
low effort0.3 %independentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
high effort0.2 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
high effort0.2 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
high effort0.1 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=public_eval · model_type=CoT
high effort0.1 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
medium effort0.1 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
medium effort0.1 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
medium effort0.0 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=public_eval · model_type=CoT
medium effort0.0 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
low effort0.0 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
low effort0.0 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=public_eval · model_type=CoT
low effort0.0 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
low effort0.0 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
minimal effort0.0 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
minimal effort0.0 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
minimal effort0.0 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=public_eval · model_type=CoT
minimal effort0.0 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
minimal effort0.0 %independentARC Prize Leaderboard
EnigmaEval
contamination=Potential contamination warning: This model was evaluated after the public release of EnigmaEval, allowing model builder access to the prompts and solutions.
8.2 %
6.6–9.8
independentScale Labs
FrontierMathhigh effort27.2 %independentEpoch AI Benchmarking Hub
2025-11-13
FrontierMathmedium effort20.3 %independentEpoch AI Benchmarking Hub
2025-11-13
FrontierMath Tier 4older versionhigh effort6.3 %independentEpoch AI Benchmarking Hub
2025-10-30
FrontierMath Tier 4older versionmedium effort4.2 %independentEpoch AI Benchmarking Hub
2025-08-07
MATH-500
implementation=vals-ai
high effort94.8 %independentVals AI
MATH Level 5high effort97.8 %independentEpoch AI Benchmarking Hub
2025-10-30
MATH Level 5medium effort96.8 %independentEpoch AI Benchmarking Hub
2025-08-20
OTIS Mock AIME 2024–2025high effort86.7 %independentEpoch AI Benchmarking Hub
2025-10-30
OTIS Mock AIME 2024–2025medium effort78.3 %independentEpoch AI Benchmarking Hub
2025-08-07

Agent + model results

systems, not bare-model scores
agent + model Epoch Inspect harness + GPT 5 Mini (2025-08-07)SWE-bench Verified64.7 %independentEpoch AI Benchmarking Hub
agent + model mini-SWE-agent + GPT 5 Mini (2025-08-07)SWE-bench Verified59.8 %communitySWE-bench Leaderboard
agent + model mini-SWE-agent + GPT 5 Mini (2025-08-07)SWE-bench bash-only59.8 %communitySWE-bench Leaderboard
agent + model mini-SWE-agent + GPT 5 Mini (2025-08-07)SWE-bench bash-only56.2 %unverifiedSWE-bench Leaderboard
agent + model mini-SWE-agent + GPT 5 Mini (2025-08-07)SWE-bench Verified56.2 %unverifiedSWE-bench Leaderboard
agent + model mini-SWE-agent + GPT 5 Mini (2025-08-07)SWE-bench Multilingual39.7 %communitySWE-bench Leaderboard
agent + model mini-SWE-agent + GPT 5 Mini (2025-08-07)SWE-bench Multilingual0.1 usd_per_taskcommunitySWE-bench Leaderboard
agent + model mini-SWE-agent + GPT 5 Mini (2025-08-07)SWE-bench bash-only0.0 usd_per_taskunverifiedSWE-bench Leaderboard
agent + model mini-SWE-agent + GPT 5 Mini (2025-08-07)SWE-bench Verified0.0 usd_per_taskunverifiedSWE-bench Leaderboard
agent + model mini-SWE-agent + GPT 5 Mini (2025-08-07)SWE-bench Verified0.0 usd_per_taskcommunitySWE-bench Leaderboard
agent + model mini-SWE-agent + GPT 5 Mini (2025-08-07)SWE-bench bash-only0.0 usd_per_taskcommunitySWE-bench Leaderboard
agent + model spoox-m + GPT 5 Mini (2025-08-07)Terminal-Bench 2.034.8 %unverifiedTerminal-Bench Leaderboard
agent + model Artificial Analysis harness + GPT 5 Mini (2025-08-07)Terminal-Bench Hard33.3 %independentArtificial Analysis
agent + model Codex CLI + GPT 5 Mini (2025-08-07)Terminal-Bench 2.031.9 %communityTerminal-Bench Leaderboard
agent + model terminus-2 + GPT 5 Mini (2025-08-07)Terminal-Bench 1.030.8 %communityTerminal-Bench Leaderboard
agent + model OpenHands + GPT 5 Mini (2025-08-07)Terminal-Bench 2.029.2 %communityTerminal-Bench Leaderboard
agent + model Artificial Analysis harness + GPT 5 Mini (2025-08-07)Terminal-Bench Hard28.8 %independentArtificial Analysis
agent + model terminus-2 + GPT 5 Mini (2025-08-07)Terminal-Bench 2.024.0 %communityTerminal-Bench Leaderboard
agent + model mini-SWE-agent + GPT 5 Mini (2025-08-07)Terminal-Bench 2.022.3 %communityTerminal-Bench Leaderboard
agent + model Artificial Analysis harness + GPT 5 Mini (2025-08-07)Terminal-Bench Hard14.4 %independentArtificial Analysis
agent + model Artificial Analysis harness + GPT 5 Mini (2025-08-07)Terminal-Bench 2.13.8 %independentArtificial Analysis

These scores measure the whole agent system (scaffold, tools, budgets) — they are never merged into the bare model’s numbers.