Opus-4.8
Anthropic
Parameters
—
Architecture
—
Released
—
License
—
About
Opus-4.8 - proprietary model by Anthropic, used as comparison model in DeepSeek-V4-Pro-0813 benchmark table.
Benchmark Scores
| Benchmark | Score | Date | Standing |
|---|---|---|---|
|
ExploitGym (6h)
cybersecurity
|
36.47%
|
28.08.2026 |
#4/8
better than 57%
|
|
Terminal Bench 2.1
coding_agent
|
93.79%
|
31.07.2026 |
#9/44
better than 81%
|
|
NL2Repo
coding_agent
|
91.38%
|
31.07.2026 |
#2/41
better than 95%
|
|
Cybergym
general_agent
|
89.88%
|
31.07.2026 |
#6/17
better than 62%
|
|
DeepSWE
coding_agent
|
79.78%
|
31.07.2026 |
#11/21
better than 50%
|
|
Toolathlon Verified
general_agent
|
96.61%
|
31.07.2026 |
#4/35
better than 88%
|
|
Automation-Bench
general_agent
|
31.54%
|
31.07.2026 |
#19/27
better than 27%
|
|
DSBench-FullStack
coding_agent
|
86.07%
|
31.07.2026 |
#3/8
better than 71%
|
|
DSBench-Hard
coding_agent
|
100.00%
|
31.07.2026 |
#1/9
better than 100%
|
|
HLE (with tools)
stem_reasoning
|
78.43%
|
13.08.2026 |
#11/49
better than 77%
|
|
Humanity's Last Exam
stem_reasoning
|
90.34%
|
13.08.2026 |
#4/71
better than 96%
|
|
Terminal-Bench 3.0
coding_agent
|
55.00%
|
28.08.2026 |
#5/7
better than 33%
|
|
ProgramBench
coding_agent
|
18.14%
|
28.08.2026 |
#15/18
better than 12%
|
|
FrontierSWE
coding_agent
|
63.34%
|
28.08.2026 |
#8/12
better than 36%
|
|
SWE-Marathon
coding_agent
|
94.28%
|
28.08.2026 |
#3/17
better than 88%
|
|
PostTrainBench
coding_agent
|
67.58%
|
28.08.2026 |
#14/22
better than 38%
|
|
ExploitGym (2h)
cybersecurity
|
32.67%
|
28.08.2026 |
#4/7
better than 50%
|
|
ExploitBench
cybersecurity
|
29.10%
|
28.08.2026 |
#5/8
better than 43%
|
|
GDPVal-AA v2
general_agent
|
86.73%
|
28.08.2026 |
#13/38
better than 68%
|
|
Chartography
multimodal_agent
|
3.74%
|
08.09.2026 |
#3/4
better than 33%
|
|
Agents' Last Exam
general_agent
|
46.32%
|
31.07.2026 |
#14/29
better than 50%
|
|
DeepSWE 1.1
coding_agent
|
75.26%
|
28.08.2026 |
#19/34
better than 45%
|
|
ApexBench
multimodal_agent
|
28.76%
|
08.09.2026 |
#2/4
better than 67%
|
|
ZEROBench
vision_language
|
64.58%
|
08.09.2026 |
#6/14
better than 62%
|
What is a good ExploitGym (6h) score? What is a good Terminal Bench 2.1 score? What is a good NL2Repo score?
No comments yet
Be the first to share your thoughts!