‹ PublicAI Index
The LLM benchmark aggregator.
GPT-5.6 Luna
OpenAI
Strongest in Cybersecurity (#11 of 47), weakest in Toxicity avoidance (#196 of 268). Above par in 28 of 37 scopes. Among the models it meets almost everywhere, it finishes behind Claude Fable 5.1 and GPT-5.6 Sol and ahead of GLM-5.3 Flash and GPT-5 mini.
§ 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.
Agents57−9.4#35/2982/6
Knowledge work61.2−9.5#39/2082/2
Tool use57.1−16.9—/81✱0/1
Web research55.3−8.6—/35✱0/2
Workflow automation55.9−10.1—/0✱✱
Knowledge55.1−7.4#39/1972/3
Document parsing55.1−8.2#35/831/1
Academic knowledge56−7.9#49/1231/1
Factuality50.7−10.1—/0✱✱
Coding54.4−11.8#44/1793/6
Agentic coding54.6−12.3#42/1733/5
Code generation63.8−1.5—/20✱0/1
Human preference61.8−7#63/3541/1
Human preference61.8−7#63/3541/1
Professional52.7−9.7#64/1941/2
Cybersecurity57.8−6.1#11/471/1
Finance57.7−4.7#22/1621/1
IT operations40.2−33.6#28/331/1
Biology research38−32.7#28/311/1
Legal54.3−11.6#51/1731/1
Medical52.7−13.8#66/1491/1
Safety53.7−7.7#78/3291/3
Secure code62.3−3.9#40/2701/1
Harm refusal54.1−7.4#86/2921/2
Fairness49.9−20.1#117/2921/2
Jailbreak resistance53.8−11.2#131/2681/1
Toxicity avoidance50−9#196/2681/1
Reasoning51.7−13.7#86/1954/5
Science60.9−2.6#19/1221/1
Reasoning50.4−18.6#63/1472/3
Mathematics48.4−15.4#93/1522/2
Expert reasoning62−4.4—/0✱✱
Core abilities43.4−21.9#161/1992/4
Language42.8−23.8#52/621/1
Instruction following36−37.8#54/621/1
General intelligence48.2−21.2#105/1781/3
Long context62.2−2.4—/0✱✱
§ 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-5.6 Luna, left for the other.
§ 3 · Sources
Where the numbers come from
12 publications, 61 figures. Every one links to the page it was read from.
LMArena Text 1453
GDPval-AA 1468
AA-Briefcase 1339
Terminal-Bench 17.6% (Codex)
ARC-AGI-2 59.6%
LiveBench 1.0 73.6LiveBench 1.0 · Reasoning 85.6LiveBench 1.0 · Coding 82.9LiveBench 1.0 · Agentic Coding 48.4LiveBench 1.0 · Mathematics 87.2LiveBench 1.0 · Data Analysis 78LiveBench 1.0 · Language 72.6LiveBench 1.0 · Instruction Following 60.1
Kagi LLM Benchmark 49.1%
SimpleBench 46.8%
Vals · Legal Research Bench 36.54%Vals · LegalBench 84.03%Vals · Finance Agent 55.04%Vals · CorpFin 64.22%Vals · TaxEval 76.17%Vals · MortgageTax 67.29%Vals · MedCode 42.39%Vals · MedScribe 84.39%Vals · BioMysteryBench 61.48%Vals · CyberBench 73.63%Vals · SRE Bench 1.53%Vals · SWE-bench Verified 93% (Mini-SWE-agent)Vals · Vibe Code Bench 77.06% (OpenHands)Vals · Code Migration 44.55%Vals · GPQA Diamond 91.67%Vals · MMLU Pro 86.04%Vals · ProofBench 60%
ParseBench · Overall 66.73%ParseBench · Tables 88.3%ParseBench · Charts 49.16%ParseBench · Content faithfulness 86.4%ParseBench · Semantic formatting 64.95%ParseBench · Visual grounding 44.83%
Enkrypt · Jailbreak risk 9.6%Enkrypt · Harmful content risk 0%Enkrypt · CBRN risk 22.2%Enkrypt · Toxicity risk 6.2%Enkrypt · Bias risk 78%Enkrypt · Insecure code risk 8%
GDPVal-AA 1569tau3-Banking 31.1%Toolathlon Verified 67.5%Automation Bench Public 33.5%Apex-Agents (pass@1) 28.6%MCPMark 66.9%BrowseComp 83.3%WildClawBench 50.4%Terminal-Bench 2.1 80.9%SciCode 52.5%SWE Bench Pro 48.8%Humanity's Last Exam (without tools) 39.5%GPQA Diamond 91.1%CritPt 21%AA-LCR 78.3%AA-Omniscience Accuracy 43%AA-Omniscience Non-Hallucination 7%
Badge
[](https://publicai.io/model-index/m/gpt-5-6-luna)