‹ PublicAI Index
The LLM benchmark aggregator.
Claude Fable 5.1
Anthropic
Strongest in Academic knowledge (#1 of 123), weakest in Cybersecurity (#22 of 47). Above par in 28 of 28 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 5.5 and Gemini 4 Argon and ahead of Claude Fable 5 and GPT-6.1 Sol.
§ 1 · Profile
What it is good at
Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.
Knowledge62.4−0.1#2/1972/3
Academic knowledge63.9leads#1/1231/1
Document parsing62.6−0.7#3/831/1
Reasoning64.2−1.2#3/1955/5
Mathematics63.7−0.1#2/1522/2
Reasoning65.5−3.5#7/1473/3
Science62.1−1.4#9/1221/1
Agents64.2−2.2#3/2983/6
Long-horizon work70.2−3.8#3/311/1
Knowledge work68.1−2.6#6/2082/2
Web research57.9−6#8/351/2
Computer use61.5−6.6#12/681/2
Coding63.4−2.8#3/1794/6
Agentic coding63.9−3#3/1734/5
Professional60.3−2.1#4/1942/2
Document design63.3−5.4#3/311/1
Medical62.6−3.9#4/1491/1
Legal62.6−3.3#6/1731/1
Finance59.6−2.8#6/1621/1
IT operations56.4−17.4#10/331/1
Cybersecurity54.9−9#22/471/1
Human preference66.6−2.2#7/3541/1
Human preference66.6−2.2#7/3541/1
Core abilities61.2−4.1#10/1992/4
Language65−1.6#3/621/1
Data analysis58.9−6.3#7/311/1
Instruction following58.9−14.9#16/621/1
General intelligence61.7−7.7#20/1781/3
§ 2 · Head to head
What it beats, and what beats it
The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for Claude Fable 5.1, left for the other.
§ 3 · Sources
Where the numbers come from
10 publications, 42 figures. Every one links to the page it was read from.
LMArena Text 1503
GDPval-AA 1763
AA-Briefcase 1669
Terminal-Bench 57.6% (Claude Code)
ARC-AGI-2 90%
LiveBench 84LiveBench · Agentic 74.2LiveBench · Reasoning 91.7LiveBench · Coding 89.7LiveBench · Browser Use 89.8LiveBench · Data Analysis 80.3LiveBench · Research 88.5LiveBench · Design 81.6
LiveBench 1.0 83.4LiveBench 1.0 · Reasoning 91.7LiveBench 1.0 · Coding 86.4LiveBench 1.0 · Agentic Coding 66.1LiveBench 1.0 · Mathematics 97LiveBench 1.0 · Data Analysis 80.3LiveBench 1.0 · Language 89.5LiveBench 1.0 · Instruction Following 73
SimpleBench 86.6%
Vals · Legal Research Bench 55.29%Vals · LegalBench 88.51%Vals · Finance Agent 58.88%Vals · TaxEval 75.96%Vals · MortgageTax 70.79%Vals · MedCode 53.51%Vals · MedScribe 91.29%Vals · CyberBench 70.42%Vals · SRE Bench 22.9%Vals · Vibe Code Bench 90.26% (OpenHands)Vals · Code Migration 54.61%Vals · GPQA Diamond 93.43%Vals · MMLU Pro 92.38%Vals · ProofBench 100%
ParseBench · Overall 77.81%ParseBench · Tables 91.53%ParseBench · Charts 66.93%ParseBench · Content faithfulness 90.77%ParseBench · Semantic formatting 76.02%ParseBench · Visual grounding 63.81%
Badge
[](https://publicai.io/model-index/m/claude-fable-5-1)