‹ PublicAI Index
The LLM benchmark aggregator.
GPT-5.5
OpenAI
Strongest in General intelligence (#5 of 178, on 1 of its 3 boards), weakest in Toxicity avoidance (#214 of 268). Above par in 27 of 29 scopes. Among the models it meets almost everywhere, it finishes behind Claude Fable 5.1 and Claude Fable 5 and ahead of Muse Spark 1.1 and Grok 4.7.
§ 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 abilities61.5−3.8#7/1992/4
General intelligence67.6−1.8#5/1781/3
Language62.3−4.3#8/621/1
Instruction following54.8−19#25/621/1
Reasoning60.2−5.2#13/1954/5
Science61.9−1.6#10/1221/1
Mathematics58.8−5#16/1521/2
Reasoning60.2−8.8#19/1472/3
Professional56.4−6#20/1941/2
Finance58.2−4.2#12/1621/1
Legal59−6.9#18/1731/1
IT operations42−31.8#20/331/1
Medical57.9−8.6#22/1491/1
Human preference64.5−4.3#26/3541/1
Human preference64.5−4.3#26/3541/1
Knowledge55.7−6.8#32/1972/3
Academic knowledge58.6−5.3#21/1231/1
Document parsing55.4−7.9#34/831/1
Coding55.3−10.9#32/1792/6
Agentic coding55.5−11.4#34/1732/5
Safety55.2−6.2#51/3292/3
Secure code63.6−2.6#20/2701/1
Fairness60.5−9.5#39/2921/2
Factual grounding51.6−19.1#46/1041/1
Harm refusal54.5−7#79/2921/2
Jailbreak resistance54.3−10.7#125/2681/1
Toxicity avoidance47.3−11.7#214/2681/1
Agents54.8−11.6#60/2982/6
Knowledge work57.7−13#56/2082/2
§ 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 GPT-5.5, left for the other.
§ 3 · Sources
Where the numbers come from
11 publications, 42 figures. Every one links to the page it was read from.
LMArena Text 1481
GDPval-AA 1359
AA-Briefcase 1135
ARC-AGI-2 85%
LiveBench 1.0 80.2LiveBench 1.0 · Reasoning 89.7LiveBench 1.0 · Coding 82.1LiveBench 1.0 · Agentic Coding 54LiveBench 1.0 · Mathematics 95.9LiveBench 1.0 · Data Analysis 81.6LiveBench 1.0 · Language 87.4LiveBench 1.0 · Instruction Following 70.7
Kagi LLM Benchmark 88.8%
SimpleBench 69%
Vals · Legal Research Bench 40.38%Vals · CaseLaw 66.24%Vals · LegalBench 86.52%Vals · Finance Agent 51.76%Vals · CorpFin 68.42%Vals · TaxEval 74.98%Vals · MortgageTax 68.76%Vals · MedCode 49.1%Vals · MedScribe 86.87%Vals · SRE Bench 3.82%Vals · SWE-bench Verified 82.6% (Mini-SWE-agent)Vals · Vibe Code Bench 69.85% (OpenHands)Vals · Code Migration 45.16%Vals · GPQA Diamond 93.18%Vals · MMLU Pro 88.14%
ParseBench · Overall 66.42%ParseBench · Tables 90.11%ParseBench · Charts 64.84%ParseBench · Content faithfulness 88.4%ParseBench · Semantic formatting 58.09%ParseBench · Visual grounding 30.64%
Enkrypt · Jailbreak risk 9.2%Enkrypt · Harmful content risk 0.6%Enkrypt · CBRN risk 20.7%Enkrypt · Toxicity risk 8%Enkrypt · Bias risk 61.5%Enkrypt · Insecure code risk 5.3%
Vectara · Factual consistency 90.7%
Badge
[](https://publicai.io/model-index/m/gpt-5-5)