Request a pilot
‹ PublicAI Index

The LLM benchmark aggregator.

GPT-5.6 Luna

OpenAI

Strongest in Decisions (#3 of 85), weakest in Toxicity avoidance (#198 of 272). Above par in 31 of 38 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 5.5 and Claude Fable 5 and ahead of Gemini 3.6 Flash and GPT-4.1.

§ 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.

Decisions70.6#3/85
Routing & classification74#3/85
Calibration67.2#4/85
Coding57#23/165
Code generation61.3#7/77
Agentic coding55.9#34/157
Agents60.5#26/268
Knowledge work63.7#23/178
Tool use57.1—/81✱
Web research57.4—/4✱
Workflow automation54.4—/0✱
Professional52.8#52/168
Biology research35.9#15/16
Finance58#19/151
Medical53.2#60/140
Legal51.5#67/151
Knowledge55#53/138
Academic knowledge56#49/123
Factuality50.6—/0✱
Human preference62.2#57/342
Human preference62.2#57/342
Reasoning52.7#70/178
Science60.9#19/122
Reasoning52.5#51/140
Mathematics49.3#82/140
Expert reasoning58—/0✱
Safety53.7#77/337
Secure code62.6#39/274
Harm refusal54#87/300
Fairness49.9#118/300
Jailbreak resistance53.8#133/272
Toxicity avoidance50.1#198/272
Core abilities45.3#143/204
Data analysis57.7#21/57
Instruction following36.3#49/57
Language37.4#50/57
General intelligence47.6#118/204
Long context62.2—/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.

GPT-4.1 Nanowon 24 · lost 1
GPT-4o Miniwon 22 · lost 2
Llama 4 Scout Instructwon 21 · lost 2
Command Awon 21 · lost 2
Grok 4.3won 25 · lost 3
Gemma 4 31Bwon 24 · lost 3
DeepSeek V3won 22 · lost 3
GPT-5.4 Nanowon 20 · lost 3
GPT-4.1 Miniwon 21 · lost 4
Gemini 2.5 Flashwon 20 · lost 5
Kimi K2.5won 19 · lost 5
GLM-4.6won 18 · lost 6
Mimo V2.5 Prowon 17 · lost 6
MiniMax M3won 25 · lost 9
DeepSeek V3.2won 18 · lost 7
Kimi K2.6won 20 · lost 8
GPT OSS 120Bwon 17 · lost 7
Kimi K2won 17 · lost 7
O3 Miniwon 16 · lost 7
Qwen3.8 27Bwon 18 · lost 8
Claude Haiku 4.5won 17 · lost 8
Grok 4won 16 · lost 8
DeepSeek R1won 16 · lost 8
DeepSeek V4 Flashwon 20 · lost 11
Claude Sonnet 4won 16 · lost 9
Gemini 2.5 Prowon 16 · lost 9
GPT-6 Lunawon 18 · lost 11
Inklingwon 21 · lost 13
Qwen3.7 Maxwon 17 · lost 11
O4 Miniwon 15 · lost 10
Qwen3 235B A22B Instructwon 14 · lost 10
GLM-5.2won 19 · lost 14
GPT-4.1won 13 · lost 11
Gemini 3.6 Flashwon 15 · lost 14
Gemini 3.5 Flashwon 14 · lost 15
GPT-5 Miniwon 12 · lost 13
Claude Opus 4won 11 · lost 12
GLM-5.3 Flashwon 13 · lost 15
Claude Sonnet 4.6won 12 · lost 14
O3won 10 · lost 15
DeepSeek V4 Prowon 12 · lost 18
Grok 4.5won 11 · lost 18
GPT-5won 9 · lost 15
Claude Sonnet 4.5won 9 · lost 16
Claude Opus 4.5won 8 · lost 15
Gemini 3.1 Prowon 8 · lost 15
GPT-6 Solwon 9 · lost 17
GPT-5.2won 9 · lost 19
GPT-5.4won 9 · lost 19
Gemini 3.8 Flashwon 9 · lost 20
Grok 4.7won 8 · lost 18
DeepSeek V4.1 Flashwon 8 · lost 19
GLM-5.3won 7 · lost 17
Claude Opus 4.6won 7 · lost 17
Qwen3.8 Maxwon 8 · lost 20
Muse Spark 1.1won 8 · lost 20
Claude Opus 4.7won 7 · lost 21
GPT-5.6 Terrawon 8 · lost 25
Grok 4.6won 7 · lost 22
Claude Sonnet 5won 7 · lost 27
GPT-5.6 Solwon 6 · lost 26
Claude Opus 4.8won 5 · lost 23
Claude Opus 5won 5 · lost 26
GPT-5.5won 4 · lost 24
Kimi K3won 3 · lost 22
Claude Fable 5won 0 · lost 23
Claude Opus 5.5won 0 · lost 26

§ 3 · Sources

Where the numbers come from

12 publications, 56 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-09-26
LMArena Text 1454
GDPval-AA ↗1 measureread 2026-09-26
GDPval-AA 1443
AA-Briefcase ↗1 measureread 2026-09-26
AA-Briefcase 1342
Terminal-Bench ↗1 measureread 2026-09-26
Terminal-Bench 17.3% (Codex)
ARC-AGI-2 ↗1 measureread 2026-09-26
ARC-AGI-2 59.6%
LiveBench ↗8 measuresread 2026-09-26
LiveBench 73.6LiveBench · Reasoning 85.6LiveBench · Coding 82.9LiveBench · Agentic Coding 48.4LiveBench · Mathematics 87.2LiveBench · Data Analysis 78LiveBench · Language 72.6LiveBench · Instruction Following 60.1
Kagi LLM Benchmark ↗1 measureread 2026-09-26
Kagi LLM Benchmark 49.1%
SimpleBench ↗1 measureread 2026-09-26
SimpleBench 46.8%
Vals.ai ↗16 measuresread 2026-09-26
Vals · Legal Research Bench 36.54%Vals · LegalBench 84.03%Vals · Harvey Legal Agent Benchmark 1.25%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 · 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%
JevBench ↗2 measuresread 2026-09-26
JevBench · Intelligence 93.1%JevBench · Calibration 87.4%
Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-09-26
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

PublicAI Index badge for GPT-5.6 Luna[![PublicAI Index](https://publicai.io/model-index/badge?model=gpt-5-6-luna)](https://publicai.io/model-index/m/gpt-5-6-luna)

✱ Placed by a one-off publication, not a board that re-ran the model. ✱✱ marks a figure the model’s own publisher printed, which counts for less again. Scores are 0–100 on the PublicAI Index scale, 50 = the average of the models each source lists. Scores belong to their publishers.

Snapshot 2026-09-26 · the full index · JSON API