‹ PublicAI Index
The LLM benchmark aggregator.
Claude Sonnet 5.5
Anthropic
Strongest in Biology research (#1 of 28), weakest in Legal (#47 of 164). Above par in 25 of 27 scopes. Among the models it meets almost everywhere, it finishes behind Claude Fable 5.1 and Claude Opus 5.5 and ahead of GPT-6 Astra 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.
Agents65.7−0.2#3/2893/6
Knowledge work71.7−0.1#2/1982/2
Long-horizon work72.8−1.2#4/481/1
Web research59−3.6#8/491/2
Computer use59−8.7#17/811/2
Professional58.8−4.3#6/1882/2
Biology research68.2leads#1/281/1
Document design64.7−2.6#2/481/1
Medical62.2−4.3#5/1491/1
IT operations61.7−12.1#8/311/1
Finance56.8−5.7#30/1611/1
Cybersecurity45.4−18.6#35/461/1
Legal54.2−11.7#47/1641/1
Reasoning62.8−2.1#6/1905/5
Mathematics63.7−0.4#3/1502/2
Reasoning62.9−3.8#11/1443/3
Coding59.1−7.3#15/1764/6
Agentic coding59.4−7.6#14/1694/5
Knowledge57.7−5#17/1911/3
Document parsing58.2−5.4#15/771/1
Human preference64.1−4.5#27/3501/1
Human preference64.1−4.5#27/3501/1
Core abilities55.9−9.3#40/2253/4
General intelligence67.1−1.7#6/2142/3
Language57−9.6#17/621/1
Instruction following54.5−19.3#26/621/1
Data analysis33.7−32.4#42/481/1
§ 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 Sonnet 5.5, left for the other.
§ 3 · Sources
Where the numbers come from
11 publications, 40 figures. Every one links to the page it was read from.
LMArena Text 1477
Artificial Analysis Intelligence Index 56
GDPval-AA 1838
AA-Briefcase 1817
Terminal-Bench 61.8% (Claude Code)
ARC-AGI-2 83.9%
LiveBench 82.5LiveBench · Agentic 78.8LiveBench · Reasoning 91.6LiveBench · Coding 94.7LiveBench · Browser Use 84LiveBench · Data Analysis 59.5LiveBench · Research 83.6LiveBench · Design 85
LiveBench 1.0 77.8LiveBench 1.0 · Reasoning 86.8LiveBench 1.0 · Coding 88.9LiveBench 1.0 · Agentic Coding 39.3LiveBench 1.0 · Mathematics 96.7LiveBench 1.0 · Data Analysis 78.6LiveBench 1.0 · Language 83.4LiveBench 1.0 · Instruction Following 70.5
SimpleBench 75.9%
Vals · Legal Research Bench 48.08%Vals · Harvey Legal Agent Benchmark 10.42%Vals · Finance Agent 58.1%Vals · MedCode 52.92%Vals · MedScribe 91.1%Vals · BioMysteryBench 81.11%Vals · CyberBench 59.58%Vals · SRE Bench 30.15%Vals · Vibe Code Bench 92.39% (OpenHands)Vals · Code Migration 69.83%Vals · ProofBench 100%
ParseBench · Overall 70.18%ParseBench · Tables 91.04%ParseBench · Charts 37.69%ParseBench · Content faithfulness 91.64%ParseBench · Semantic formatting 71.63%ParseBench · Visual grounding 58.92%
Badge
[](https://publicai.io/model-index/m/claude-sonnet-5-5)