Fable-5
Anthropic
Parameters
—
Architecture
—
Released
—
License
—
About
Fable-5 (w/ fallback) - proprietary model by Anthropic, used as comparison model in DeepSeek-V4-Pro-0813 benchmark table.
Benchmark Scores
| Benchmark | Score | Date | Standing |
|---|---|---|---|
|
Terminal Bench 2.1
coding_agent
|
93.35%
|
18.08.2026 |
#10/44
better than 77%
|
|
SWE-bench Pro
coding_agent
|
89.19%
|
18.08.2026 |
#2/77
better than 99%
|
|
MLS-Bench-Lite
coding_agent
|
100.00%
|
18.08.2026 |
#1/9
better than 100%
|
|
PaperBench
coding_agent
|
93.27%
|
18.08.2026 |
#3/10
better than 78%
|
|
AndroidBench
coding_agent
|
100.00%
|
18.08.2026 |
#1/5
better than 100%
|
|
QwenSWEBench
coding_agent
|
100.00%
|
18.08.2026 |
#1/9
better than 100%
|
|
QwenQoderBench
coding_agent
|
100.00%
|
18.08.2026 |
#1/5
better than 100%
|
|
QwenReactBench
coding_agent
|
100.00%
|
18.08.2026 |
#1/5
better than 100%
|
|
QwenSVGBench
coding_agent
|
73.75%
|
18.08.2026 |
#3/5
better than 50%
|
|
CoWorkBench
general_agent
|
100.00%
|
18.08.2026 |
#1/12
better than 100%
|
|
JobBench
general_agent
|
77.06%
|
18.08.2026 |
#5/19
better than 78%
|
|
SkillsBench
general_agent
|
93.67%
|
18.08.2026 |
#2/9
better than 88%
|
|
Automation-Bench
general_agent
|
35.19%
|
18.08.2026 |
#16/27
better than 42%
|
|
Toolathlon Verified
general_agent
|
100.00%
|
18.08.2026 |
#1/35
better than 100%
|
|
WideSearch
general_agent
|
90.02%
|
18.08.2026 |
#8/24
better than 70%
|
|
GPQA Diamond
stem_reasoning
|
96.79%
|
18.08.2026 |
#10/80
better than 87%
|
|
Humanity's Last Exam
stem_reasoning
|
96.97%
|
18.08.2026 |
#2/71
better than 99%
|
|
$OneMillion-Bench
general_capabilities
|
53.61%
|
18.08.2026 |
#6/11
better than 50%
|
|
PLawBench
general_capabilities
|
79.02%
|
18.08.2026 |
#3/5
better than 50%
|
|
PRBench-Legal
general_capabilities
|
100.00%
|
18.08.2026 |
#1/5
better than 50%
|
|
PRBench-Finance
general_capabilities
|
78.26%
|
18.08.2026 |
#2/5
better than 75%
|
|
ExploitGym (6h)
cybersecurity
|
83.11%
|
28.08.2026 |
#2/8
better than 86%
|
|
HLE (with tools)
stem_reasoning
|
90.20%
|
18.08.2026 |
#2/49
better than 98%
|
|
SWE-Marathon
coding_agent
|
63.31%
|
28.08.2026 |
#8/17
better than 56%
|
|
Cybergym
general_agent
|
89.88%
|
13.08.2026 |
#6/17
better than 62%
|
|
DeepSWE
coding_agent
|
96.29%
|
13.08.2026 |
#2/21
better than 95%
|
|
DSBench-FullStack
coding_agent
|
100.00%
|
13.08.2026 |
#1/8
better than 100%
|
|
DSBench-Hard
coding_agent
|
92.59%
|
13.08.2026 |
#2/9
better than 88%
|
|
Terminal-Bench 3.0
coding_agent
|
97.00%
|
28.08.2026 |
#2/7
better than 83%
|
|
ProgramBench
coding_agent
|
43.54%
|
28.08.2026 |
#8/18
better than 59%
|
|
FrontierSWE
coding_agent
|
100.00%
|
28.08.2026 |
#1/12
better than 100%
|
|
PostTrainBench
coding_agent
|
97.95%
|
28.08.2026 |
#2/22
better than 95%
|
|
ExploitGym (2h)
cybersecurity
|
82.67%
|
28.08.2026 |
#2/7
better than 83%
|
|
ExploitBench
cybersecurity
|
100.00%
|
28.08.2026 |
#1/8
better than 100%
|
|
GDPVal-AA v2
general_agent
|
95.19%
|
28.08.2026 |
#4/38
better than 92%
|
|
WorkSpaceBench
general_agent
|
62.50%
|
18.08.2026 |
#2/11
better than 90%
|
|
Agents' Last Exam
general_agent
|
40.49%
|
28.08.2026 |
#19/29
better than 32%
|
|
DeepSWE 1.1
coding_agent
|
90.89%
|
18.08.2026 |
#7/34
better than 79%
|
|
IFBench
instruction_following
|
66.34%
|
18.08.2026 |
#32/42
better than 24%
|
What is a good Terminal Bench 2.1 score? What is a good SWE-bench Pro score? What is a good MLS-Bench-Lite score?
No comments yet
Be the first to share your thoughts!