Claude Opus 5
Anthropic
Parameters
—
Architecture
—
Released
—
License
—
Openness Index Score
25.0/100
About
Proprietary Anthropic frontier model; included as comparison model in the Tencent Hy4 preview benchmark appendix.
Benchmark Scores
| Benchmark | Score | Date | Standing |
|---|---|---|---|
|
Terminal-Bench 4.0
coding_agent
|
73.82%
|
19.10.2026 |
#2/10
better than 89%
|
|
OfficeQA Pro
general_agent
|
98.99%
|
28.08.2026 |
#2/14
better than 92%
|
|
Apex-Agents
general_agent
|
56.39%
|
28.08.2026 |
#3/24
better than 91%
|
|
HLE (with tools)
stem_reasoning
|
83.78%
|
28.08.2026 |
#6/49
better than 90%
|
|
SWE-Marathon
coding_agent
|
96.65%
|
28.08.2026 |
#2/17
better than 94%
|
|
GDP.pdf
long_context
|
58.06%
|
19.10.2026 |
#4/13
better than 75%
|
|
DeepSWE
coding_agent
|
94.64%
|
28.08.2026 |
#3/21
better than 90%
|
|
SWE Atlas - QnA
coding_agent
|
88.14%
|
28.08.2026 |
#4/24
better than 87%
|
|
SWE Atlas - TW
coding_agent
|
100.00%
|
28.08.2026 |
#1/20
better than 100%
|
|
SWE Atlas - RF
coding_agent
|
100.00%
|
28.08.2026 |
#1/14
better than 100%
|
|
Terminal-Bench 2.1 (Terminus-2)
coding_agent
|
96.90%
|
28.08.2026 |
#5/44
better than 91%
|
|
NL2Repo
coding_agent
|
100.00%
|
28.08.2026 |
#1/41
better than 100%
|
|
ProgramBench
coding_agent
|
52.98%
|
28.08.2026 |
#6/18
better than 65%
|
|
PostTrainBench
coding_agent
|
74.74%
|
28.08.2026 |
#12/22
better than 48%
|
|
Harbor-Index
coding_agent
|
100.00%
|
28.08.2026 |
#1/7
better than 100%
|
|
Hy-Backend 2.0 (Internal)
coding_agent
|
60.26%
|
28.08.2026 |
#3/8
better than 71%
|
|
Hy-SWE Max Verified (Internal)
coding_agent
|
100.00%
|
28.08.2026 |
#1/8
better than 100%
|
|
Hy-CompanyBench V2 (Internal)
general_agent
|
100.00%
|
28.08.2026 |
#1/8
better than 100%
|
|
WideSearch
general_agent
|
95.50%
|
28.08.2026 |
#2/24
better than 96%
|
|
$OneMillion-Bench
general_capabilities
|
100.00%
|
28.08.2026 |
#1/11
better than 100%
|
|
Hy-LifeSearch (Internal)
general_agent
|
70.20%
|
28.08.2026 |
#2/8
better than 86%
|
|
Hy-BrowseComp-Pro2 (Internal)
general_agent
|
100.00%
|
28.08.2026 |
#1/8
better than 100%
|
|
Toolathlon Verified
general_agent
|
97.21%
|
28.08.2026 |
#2/35
better than 94%
|
|
SkillsBench Avg5
coding_agent
|
95.18%
|
28.08.2026 |
#3/16
better than 87%
|
|
JobBench
general_agent
|
100.00%
|
28.08.2026 |
#1/19
better than 100%
|
|
WorkSpaceBench
general_agent
|
100.00%
|
28.08.2026 |
#1/11
better than 100%
|
|
GDPVal-AA v2
general_agent
|
100.00%
|
28.08.2026 |
#1/38
better than 100%
|
|
BankerToolBench
general_agent
|
100.00%
|
28.08.2026 |
#1/15
better than 100%
|
|
E-Bench (Internal)
general_agent
|
91.28%
|
28.08.2026 |
#2/8
better than 86%
|
|
E-Bench-Code (Internal)
coding_agent
|
96.84%
|
28.08.2026 |
#2/8
better than 86%
|
|
Hy-FinAgentBench (Internal)
domain_finance
|
92.59%
|
28.08.2026 |
#2/8
better than 86%
|
|
Hy-FinmodelBench v2 (Internal)
domain_finance
|
100.00%
|
28.08.2026 |
#1/8
better than 100%
|
|
BioMysteryBench
stem_reasoning
|
66.93%
|
28.08.2026 |
#4/10
better than 67%
|
|
CritPt (no tools)
stem_reasoning
|
89.91%
|
28.08.2026 |
#2/21
better than 95%
|
|
GPQA Diamond
stem_reasoning
|
98.87%
|
28.08.2026 |
#4/80
better than 96%
|
|
Humanity's Last Exam
stem_reasoning
|
100.00%
|
28.08.2026 |
#1/71
better than 100%
|
|
SUPERChem
stem_reasoning
|
100.00%
|
28.08.2026 |
#1/8
better than 100%
|
|
ArXivMath
stem_reasoning
|
71.22%
|
28.08.2026 |
#2/7
better than 83%
|
|
HorizonMath (pass@4)
stem_reasoning
|
25.28%
|
28.08.2026 |
#5/7
better than 33%
|
|
MathArena Apex 2025
stem_reasoning
|
100.00%
|
28.08.2026 |
#1/8
better than 100%
|
|
BrokenArXiv
stem_reasoning
|
100.00%
|
28.08.2026 |
#1/7
better than 100%
|
|
DRACO
general_agent
|
100.00%
|
28.08.2026 |
#1/12
better than 100%
|
|
SWE-bench Multilingual
coding_agent
|
100.00%
|
28.08.2026 |
#1/48
better than 100%
|
|
Agents' Last Exam
general_agent
|
44.48%
|
28.08.2026 |
#17/29
better than 43%
|
|
MCP-Atlas
general_agent
|
95.43%
|
28.08.2026 |
#2/54
better than 98%
|
|
DeepSWE 1.1
coding_agent
|
96.09%
|
19.10.2026 |
#4/34
better than 91%
|
|
CharXiv (RQ)
document_understanding
|
69.96%
|
19.10.2026 |
#6/34
better than 82%
|
|
LOCA-Bench (256k)
long_context
|
40.00%
|
19.10.2026 |
#4/7
better than 50%
|
|
SWE Refactor Bench
coding_agent
|
100.00%
|
19.10.2026 |
#1/6
better than 100%
|
|
NL2Repo-Bench
coding_agent
|
100.00%
|
19.10.2026 |
#1/23
better than 100%
|
|
Automation-Bench
general_agent
|
72.88%
|
28.08.2026 |
#7/27
better than 77%
|
|
SWE-bench Pro
coding_agent
|
88.29%
|
28.08.2026 |
#3/77
better than 97%
|
|
OSWorld 2.0 (Binary)
general_agent
|
100.00%
|
19.10.2026 |
#1/8
better than 100%
|
|
BabyVision
vision_language
|
100.00%
|
19.10.2026 |
#1/24
better than 100%
|
What is a good Terminal-Bench 4.0 score? What is a good OfficeQA Pro score? What is a good Apex-Agents score?
No comments yet
Be the first to share your thoughts!