‹ PublicAI Index
The LLM benchmark aggregator.
Claude Fable 5
Anthropic
Strongest in Computer use (#1 of 69, on 1 of its 2 boards), weakest in Knowledge work (#22 of 208). Above par in 23 of 23 scopes. Among the models it meets almost everywhere, it finishes behind Gemini 4 Carbon and Claude Opus 5 and ahead of GPT-6 Astra and Claude Sonnet 5.5.
§ 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.
Core abilities65.3−0.3#2/2002/4
Language66.6leads#1/621/1
General intelligence67.6−1.4#5/1801/3
Instruction following63.9−9.9#6/621/1
Professional60.5−1.9#3/1941/2
Medical62.8−3.7#3/1491/1
Finance61.1−1.3#5/1621/1
Legal60.8−5.1#10/1731/1
Human preference66.7−2.1#5/3541/1
Human preference66.7−2.1#5/3541/1
Coding62.1−4#6/1793/6
Agentic coding62.6−4.1#6/1733/5
Agents61.6−4.7#10/2984/6
Computer use68.1leads#1/691/2
Web research52−11.6#18/361/2
Knowledge work65.1−5.6#22/2082/2
Reasoning61.5−3.9#10/1953/5
Mathematics62−1.8#7/1522/2
Science61.9−1.6#11/1221/1
Reasoning60−9#20/1471/3
Knowledge58.5−4#16/1972/3
Academic knowledge62.8−1.1#3/1231/1
Document parsing58−5.3#18/831/1
§ 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, left for the other.
§ 3 · Sources
Where the numbers come from
10 publications, 36 figures. Every one links to the page it was read from.
LMArena Text 1504
GDPval-AA 1618
AA-Briefcase 1535
Terminal-Bench 44.8% (Claude Code)
ARC-AGI-2 89.2%
LiveBench 1.0 83LiveBench 1.0 · Reasoning 89.7LiveBench 1.0 · Coding 86LiveBench 1.0 · Agentic Coding 62.2LiveBench 1.0 · Mathematics 96LiveBench 1.0 · Data Analysis 80.5LiveBench 1.0 · Language 90.7LiveBench 1.0 · Instruction Following 75.8
OSWorld 86%
Kagi LLM Benchmark 91.4%
Vals · Legal Research Bench 49.52%Vals · LegalBench 88.56%Vals · Finance Agent 56.31%Vals · CorpFin 71.83%Vals · TaxEval 76.94%Vals · MortgageTax 68.92%Vals · MedCode 56.07%Vals · MedScribe 88.52%Vals · Web Search Index 48.45% (Exa)Vals · SWE-bench Verified 95% (Mini-SWE-agent)Vals · Vibe Code Bench 90.35% (OpenHands)Vals · Code Migration 55.06%Vals · GPQA Diamond 93.18%Vals · MMLU Pro 91.5%Vals · ProofBench 95%
ParseBench · Overall 69.79%ParseBench · Tables 89.79%ParseBench · Charts 52.21%ParseBench · Content faithfulness 90.02%ParseBench · Semantic formatting 72.62%ParseBench · Visual grounding 44.29%
Badge
[](https://publicai.io/model-index/m/claude-fable-5)