Anthropicmodel systemcard
Claude Fable 5
#1of 27 · general
Standing
dot = this model · shaded = the field's middle 60% · track = full fieldGeneral
95.7#1/27
Math
91.6#1/27
Coding
81.6#2/27
Health
76.6#4/26
Engineering
93.8#2/17
Analyst
80.0#3/11
Legal
72.5#2/26
Instruct
81.5#2/27
Agents
82.0#3/27
Professional
95.6#2/27
Multimodal
88.1#2/19
Hallucination
84.3#1/27
Speed
16.9#22/22
Value
71.8#1/27
Measured
≈ imputed · amber dot = lab-self-reported among the sourcesGPQA Diamond92.6%#11
HLE60.0%#1
MMLU-Pro91.5%#2
LiveBench80.6%#2
SWE-bench V95.0%#3
LiveCodeBench89.8%#3
SciCode60.2%#1
CritPt28.6%#4
Terminal-Bench82.6%#4
APEX Agents45.0%#1
GDPval62.0%#2
Arena1508#1
AA Index62.1#2
Vals Index75.1#1
Epoch Capabilities Index161.5#2
Operations
$10 / $50$ per 1M tokens, in / out
$20blended
63tokens / sec
$1.44$ per solved task
1508arena elo · 16,056 votes