BenchAtlas

Rankings / Anthropic

Claude Opus 4.6

released 2026-02-05

BenchAtlas Index

as of 2026-09-11
78.5
max effort
rank #21
8 families · 4 categories · medium
76.8
max effort
rank #24
8 families · 4 categories · medium
74.5
32K budget
rank #41
5 families · 3 categories · medium

Benchmark evidence

128 results
agentic coding
Release 2026-06-25
tasks_counted=3 · livebench_version=2026-06-25
high effort49.0 %independentLiveBench
τ²-Benchmax effort92.1 %independentArtificial Analysis
τ²-Benchhigh effort84.8 %independentArtificial Analysis
coding
Release 2026-06-25
tasks_counted=2 · livebench_version=2026-06-25
high effort78.2 %independentLiveBench
SciCodemax effort51.9 %independentArtificial Analysis
SciCodehigh effort45.7 %independentArtificial Analysis
data analysis
Release 2026-06-25
tasks_counted=3 · livebench_version=2026-06-25
high effort69.9 %independentLiveBench
external indices
Intelligence Index v4.1max effort31.9 pointsindependentArtificial Analysis
Intelligence Index v4.1high effort26.4 pointsindependentArtificial Analysis
ECI64K budget155.3 pointsindependentEpoch AI Benchmarking Hub
ECIlow effort155.3 pointsindependentEpoch AI Benchmarking Hub
ECImax effort155.3 pointsindependentEpoch AI Benchmarking Hub
ECImedium effort155.3 pointsindependentEpoch AI Benchmarking Hub
ECIhigh effort155.3 pointsindependentEpoch AI Benchmarking Hub
ECI120K budget155.3 pointsindependentEpoch AI Benchmarking Hub
ECI32K budget155.3 pointsindependentEpoch AI Benchmarking Hub
ECI155.3 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.
no reasoning96.3 %
95.9–96.7
independentScale Labs
MASK
contamination=Potential contamination warning: This model was evaluated after the public release of MASK, allowing model builder access to the prompts and solutions.
max effort85.4 %
84.9–85.9
independentScale Labs
SimpleQA Verifiedmax effort47.0 %independentEpoch AI Benchmarking Hub
2026-08-27
SimpleQA Verified32K budget46.5 %independentEpoch AI Benchmarking Hub
2026-02-06
SimpleQA Verified43.1 %independentEpoch AI Benchmarking Hub
2026-02-06
SimpleQA Verifiedmax effort41.0 %independentEpoch AI Benchmarking Hub
2026-02-13
human preference
Coding (style control)older version
arena=text · category=coding · style_control=true
1546.4
1540.9–1551.9
communityLMArena Leaderboard Dataset
Chinese (style control)older version
arena=text · category=chinese · style_control=true
1546.0
1536.2–1555.8
communityLMArena Leaderboard Dataset
Industry Software And It Services (style control)older version
arena=text · category=industry_software_and_it_services · style_control=true
1535.5
1530.6–1540.3
communityLMArena Leaderboard Dataset
Expert (style control)older version
arena=text · category=expert · style_control=true
1533.6
1525.7–1541.5
communityLMArena Leaderboard Dataset
Hard Prompts English (style control)older version
arena=text · category=hard_prompts_english · style_control=true
1529.7
1524.4–1535.1
communityLMArena Leaderboard Dataset
Hard Prompts (style control)older version
arena=text · category=hard_prompts · style_control=true
1526.7
1522.5–1530.9
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
1518.5
1512.1–1524.8
communityLMArena Leaderboard Dataset
Longer Query (style control)older version
arena=text · category=longer_query · style_control=true
1517.5
1512.5–1522.4
communityLMArena Leaderboard Dataset
Exclude Ties (style control)older version
arena=text · category=exclude_ties · style_control=true
1516.3
1511.6–1520.9
communityLMArena Leaderboard Dataset
Industry Mathematical (style control)older version
arena=text · category=industry_mathematical · style_control=true
1515.2
1505.1–1525.4
communityLMArena Leaderboard Dataset
Industry Medicine And Healthcare (style control)older version
arena=text · category=industry_medicine_and_healthcare · style_control=true
1514.5
1505.5–1523.5
communityLMArena Leaderboard Dataset
Multi Turn (style control)older version
arena=text · category=multi_turn · style_control=true
1511.1
1504.8–1517.4
communityLMArena Leaderboard Dataset
Russian (style control)older version
arena=text · category=russian · style_control=true
1508.0
1500.7–1515.4
communityLMArena Leaderboard Dataset
Math (style control)older version
arena=text · category=math · style_control=true
1506.5
1496.6–1516.4
communityLMArena Leaderboard Dataset
English (style control)older version
arena=text · category=english · style_control=true
1506.3
1501.7–1510.8
communityLMArena Leaderboard Dataset
Industry Legal And Government (style control)older version
arena=text · category=industry_legal_and_government · style_control=true
1505.2
1496.7–1513.8
communityLMArena Leaderboard Dataset
Spanish (style control)older version
arena=text · category=spanish · style_control=true
1501.0
1487.9–1514.2
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
1501.0
1495.0–1507.0
communityLMArena Leaderboard Dataset
Instruction Following (style control)older version
arena=text · category=instruction_following · style_control=true
1500.1
1495.0–1505.2
communityLMArena Leaderboard Dataset
Overall (style control)
arena=text · category=overall · style_control=true
1497.6
1494.1–1501.0
communityLMArena Leaderboard Dataset
Polish (style control)older version
arena=text · category=polish · style_control=true
1496.8
1480.6–1513.0
communityLMArena Leaderboard Dataset
French (style control)older version
arena=text · category=french · style_control=true
1494.1
1480.1–1508.2
communityLMArena Leaderboard Dataset
German (style control)older version
arena=text · category=german · style_control=true
1492.4
1474.2–1510.6
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
1491.4
1485.8–1497.1
communityLMArena Leaderboard Dataset
Non English (style control)older version
arena=text · category=non_english · style_control=true
1485.3
1481.1–1489.5
communityLMArena Leaderboard Dataset
Japanese (style control)older version
arena=text · category=japanese · style_control=true
1481.3
1458.4–1504.2
communityLMArena Leaderboard Dataset
Creative Writing (style control)older version
arena=text · category=creative_writing · style_control=true
1479.3
1472.8–1485.8
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
1478.9
1472.9–1484.8
communityLMArena Leaderboard Dataset
Korean (style control)older version
arena=text · category=korean · style_control=true
1459.3
1441.6–1477.1
communityLMArena Leaderboard Dataset
knowledge science
GPQA Diamond32K budget90.5 %independentEpoch AI Benchmarking Hub
2026-02-06
GPQA Diamond
implementation=vals-ai
max effort89.6 %independentVals AI
GPQA Diamond
implementation=artificial-analysis
max effort89.6 %independentArtificial Analysis
GPQA Diamond64K budget88.8 %independentEpoch AI Benchmarking Hub
2026-02-06
GPQA Diamondmax effort88.4 %independentEpoch AI Benchmarking Hub
2026-08-06
GPQA Diamond
implementation=artificial-analysis
high effort84.0 %independentArtificial Analysis
Humanity's Last Exam
implementation=artificial-analysis
max effort39.9 %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
max effort34.4 %
32.6–36.3
independentScale Labs
Humanity's Last Exam
implementation=artificial-analysis
high effort19.1 %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
no reasoning19.0 %
17.5–20.5
independentScale Labs
MMLU-Pro
implementation=vals-ai
max effort89.1 %independentVals AI
language
Release 2026-06-25
tasks_counted=3 · livebench_version=2026-06-25
high effort83.3 %independentLiveBench
long context instruction
AA-LCRmax effort78.0 %independentArtificial Analysis
AA-LCRhigh effort67.0 %independentArtificial Analysis
IFBenchmax effort53.1 %independentArtificial Analysis
IFBenchhigh effort44.6 %independentArtificial Analysis
Release 2026-06-25
tasks_counted=4 · livebench_version=2026-06-25
high effort63.3 %independentLiveBench
MultiChallengeno reasoning56.0 %
54.9–57.1
independentScale Labs
MultiChallengemax effort37.1 %
35.0–39.3
independentScale Labs
multimodal
MMMU
implementation=vals-ai
max effort83.9 %independentVals AI
VISTAmax effort46.1 %
45.8–46.3
independentScale Labs
VISTAno reasoning45.5 %
45.0–45.9
independentScale Labs
professional
CaseLawthinking62.1 %independentVals AI
CorpFinmax effort67.0 %independentVals AI
LegalBenchmax effort85.3 %independentVals AI
MedQAthinking95.4 %independentVals AI
TaxEvalmax effort76.0 %independentVals AI
reasoning math
AIME
implementation=vals-ai
thinking95.6 %independentVals AI
ARC-AGI-1older version
split=public_eval · model_type=CoT
max effort · 120K budget96.8 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=public_eval · model_type=CoT
high effort · 120K budget96.3 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=public_eval · model_type=CoT
medium effort · 120K budget94.8 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
high effort · 120K budget94.0 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
max effort · 120K budget93.0 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
medium effort · 120K budget92.0 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=public_eval · model_type=CoT
low effort · 120K budget89.6 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
low effort · 120K budget86.0 %independentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
high effort · 120K budget79.0 %independentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
max effort · 120K budget74.8 %independentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
medium effort · 120K budget73.6 %independentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
high effort · 120K budget69.2 %independentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
max effort · 120K budget68.8 %independentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
medium effort · 120K budget66.3 %independentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
low effort · 120K budget64.6 %independentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
low effort · 120K budget59.9 %independentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
high effort · 120K budget3.8 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
max effort · 120K budget3.8 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
max effort · 120K budget3.6 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
high effort · 120K budget3.5 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
medium effort · 120K budget3.0 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
medium effort · 120K budget2.7 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=public_eval · model_type=CoT
low effort · 120K budget2.4 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-2
split=semi_private · model_type=CoT
low effort · 120K budget2.3 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
max effort · 120K budget1.9 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
high effort · 120K budget1.4 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=public_eval · model_type=CoT
max effort · 120K budget1.3 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
medium effort · 120K budget1.0 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=public_eval · model_type=CoT
high effort · 120K budget0.9 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=semi_private · model_type=CoT
low effort · 120K budget0.6 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-1older version
split=public_eval · model_type=CoT
medium effort · 120K budget0.6 usd_per_taskindependentARC Prize Leaderboard
ARC-AGI-3older version
split=semi_private · model_type=CoT
max effort0.5 %independentARC Prize Leaderboard
ARC-AGI-1older version
split=public_eval · model_type=CoT
low effort · 120K budget0.4 usd_per_taskindependentARC 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.
max effort7.6 %
6.1–9.1
independentScale Labs
EnigmaEval
contamination=Potential contamination warning: This model was evaluated after the public release of EnigmaEval, allowing model builder access to the prompts and solutions.
no reasoning6.8 %
5.4–8.3
independentScale Labs
FrontierMathmax effort40.7 %independentEpoch AI Benchmarking Hub
2026-02-12
FrontierMath32K budget40.0 %independentEpoch AI Benchmarking Hub
2026-02-06
FrontierMath64K budget39.7 %independentEpoch AI Benchmarking Hub
2026-02-06
FrontierMath38.3 %independentEpoch AI Benchmarking Hub
2026-02-05
FrontierMath Tier 4older versionmax effort22.9 %independentEpoch AI Benchmarking Hub
2026-02-12
FrontierMath Tier 4older version64K budget20.8 %independentEpoch AI Benchmarking Hub
2026-02-06
FrontierMath Tier 4older version32K budget20.8 %independentEpoch AI Benchmarking Hub
2026-02-06
FrontierMath Tier 4older version14.6 %independentEpoch AI Benchmarking Hub
2026-02-05
Release 2026-06-25
tasks_counted=4 · livebench_version=2026-06-25
high effort89.3 %independentLiveBench
Release 2026-06-25
tasks_counted=4 · livebench_version=2026-06-25
high effort88.7 %independentLiveBench
OTIS Mock AIME 2024–202564K budget94.4 %independentEpoch AI Benchmarking Hub
2026-02-06
OTIS Mock AIME 2024–202532K budget93.1 %independentEpoch AI Benchmarking Hub
2026-02-06
OTIS Mock AIME 2024–2025max effort91.1 %independentEpoch AI Benchmarking Hub
2026-08-06

Agent + model results

systems, not bare-model scores
agent + model Epoch Inspect harness + Claude Opus 4.6SWE-bench Verified78.7 %independentEpoch AI Benchmarking Hub
agent + model Epoch Inspect harness + Claude Opus 4.6SWE-bench Verified75.6 %independentEpoch AI Benchmarking Hub
agent + model mini-SWE-agent + Claude Opus 4.6SWE-bench Verified75.6 %unverifiedSWE-bench Leaderboard
agent + model mini-SWE-agent + Claude Opus 4.6SWE-bench bash-only75.6 %unverifiedSWE-bench Leaderboard
agent + model mini-SWE-agent + Claude Opus 4.6SWE-bench Multilingual72.0 %communitySWE-bench Leaderboard
agent + model mini-SWE-agent + Claude Opus 4.6SWE-bench Multilingual0.7 usd_per_taskcommunitySWE-bench Leaderboard
agent + model mini-SWE-agent + Claude Opus 4.6SWE-bench Verified0.6 usd_per_taskunverifiedSWE-bench Leaderboard
agent + model mini-SWE-agent + Claude Opus 4.6SWE-bench bash-only0.6 usd_per_taskunverifiedSWE-bench Leaderboard
agent + model terminus-kira-env-bootstrap + Claude Opus 4.6Terminal-Bench 2.076.4 %unverifiedTerminal-Bench Leaderboard
agent + model capy-build + Claude Opus 4.6Terminal-Bench 2.075.3 %unverifiedTerminal-Bench Leaderboard
agent + model terminus-3-3 + Claude Opus 4.6Terminal-Bench 2.074.7 %unverifiedTerminal-Bench Leaderboard
agent + model judy + Claude Opus 4.6Terminal-Bench 2.071.9 %unverifiedTerminal-Bench Leaderboard
agent + model Droid + Claude Opus 4.6Terminal-Bench 2.069.9 %unverifiedTerminal-Bench Leaderboard
agent + model Crux + Claude Opus 4.6Terminal-Bench 2.066.9 %unverifiedTerminal-Bench Leaderboard
agent + model mux + Claude Opus 4.6Terminal-Bench 2.066.5 %unverifiedTerminal-Bench Leaderboard
agent + model terminus-2 + Claude Opus 4.6Terminal-Bench 2.062.9 %communityTerminal-Bench Leaderboard
agent + model Claude Code + Claude Opus 4.6Terminal-Bench 2.058.0 %communityTerminal-Bench Leaderboard
agent + model Artificial Analysis harness + Claude Opus 4.6Terminal-Bench Hard48.5 %independentArtificial Analysis
agent + model Artificial Analysis harness + Claude Opus 4.6Terminal-Bench Hard46.2 %independentArtificial Analysis

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

Related variants

Claude Opus 4 20250514Claude Opus 4.5 20251101Claude Opus 4.8Claude Opus 4.7Claude Opus 4.1 20250805Claude Opus 4.1Claude Opus 4Claude Opus 4.5