Request a pilot
‹ PublicAI Index

The LLM benchmark aggregator.

GPT-5.6 Sol

OpenAI

Strongest in Science (#2 of 122), weakest in Toxicity avoidance (#227 of 272). Above par in 32 of 37 scopes. Among the models it meets almost everywhere, it finishes behind Claude Fable 5 and Claude Opus 5.5 and ahead of Claude Opus 4.7 and Grok 4.6.

§ 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.

Reasoning62.8#6/178
Science63.4#2/122
Mathematics61.8#8/140
Reasoning63.5#10/140
Expert reasoning46.3—/0✱
Coding62.1#7/165
Code generation63.3#4/77
Agentic coding61.7#9/157
Repository Q&A44.5—/0✱
Core abilities61#7/204
Language65.8#6/57
Data analysis60.8#7/57
Instruction following56.8#19/57
General intelligence60.7#23/204
Human preference65#16/342
Human preference65#16/342
Knowledge58.2#18/138
Academic knowledge59.8#16/123
Multimodal understanding53.2—/19✱
Agents62.1#21/268
Web research55.1#2/4
Knowledge work66.9#14/178
Workflow automation48.6—/0✱
Professional56.3#26/168
IT operations60.3#3/11
Biology research53.6#7/16
Finance57.3#25/151
Legal56.1#30/151
Medical54.5#46/140
Cybersecurity60.2—/9✱
Safety53.1#101/337
Secure code63#32/274
Harm refusal55#75/300
Fairness53.7#80/300
Factual grounding43.9#82/101
Jailbreak resistance54.2#128/272
Toxicity avoidance46.3#227/272

§ 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 Sol, left for the other.

DeepSeek V4 Flashwon 32 · lost 1
Grok 4.3won 27 · lost 1
DeepSeek V3won 25 · lost 1
GPT-4.1 Nanowon 23 · lost 1
GPT-4.1 Miniwon 22 · lost 2
GPT-5.4 Nanowon 22 · lost 2
Command Awon 21 · lost 2
Grok 4won 21 · lost 2
GPT-5.4 Miniwon 21 · lost 2
Gemma 4 31Bwon 21 · lost 2
Mimo V2.5 Prowon 21 · lost 2
Kimi K2.6won 26 · lost 3
GPT-4o Miniwon 22 · lost 3
Gemini 2.5 Flashwon 22 · lost 3
DeepSeek R1won 22 · lost 3
DeepSeek V3.2won 22 · lost 3
Kimi K2.5won 22 · lost 3
Kimi K2won 21 · lost 3
GLM-4.6won 21 · lost 3
GLM-5.2won 28 · lost 4
Llama 4 Scout Instructwon 20 · lost 3
Qwen3.8 27Bwon 20 · lost 3
Qwen3.7 Maxwon 24 · lost 4
Claude Sonnet 4.6won 23 · lost 4
DeepSeek V4 Prowon 28 · lost 5
GPT-6 Lunawon 22 · lost 4
MiniMax M3won 27 · lost 5
GPT OSS 120Bwon 21 · lost 4
O4 Miniwon 21 · lost 4
Qwen3 235B A22B Instructwon 20 · lost 4
Claude Opus 4.5won 19 · lost 4
GPT-5.6 Lunawon 26 · lost 6
Claude Sonnet 4won 21 · lost 5
Gemini 2.5 Prowon 21 · lost 5
GPT-5.4won 23 · lost 6
Gemini 3.5 Flashwon 23 · lost 6
Gemini 3.6 Flashwon 23 · lost 6
DeepSeek V4.1 Flashwon 22 · lost 6
GPT-4owon 18 · lost 5
O3 Miniwon 18 · lost 5
Inklingwon 25 · lost 7
Claude Haiku 4.5won 19 · lost 6
O3won 18 · lost 6
GPT-5.6 Terrawon 22 · lost 8
GPT-5 Miniwon 18 · lost 7
Claude Sonnet 4.5won 18 · lost 7
GLM-5.3 Flashwon 20 · lost 8
Qwen3.8 Maxwon 20 · lost 8
Claude Sonnet 5won 22 · lost 9
GPT-4.1won 17 · lost 7
Claude Opus 4won 17 · lost 7
Grok 4.5won 21 · lost 9
Gemini 3.1 Prowon 16 · lost 8
GPT-6 Solwon 18 · lost 9
GLM-5.3won 17 · lost 9
GPT-5won 16 · lost 9
Claude Opus 4.6won 16 · lost 9
Gemini 3.7 Flashwon 14 · lost 9
GPT-5.2won 17 · lost 11
Muse Spark 1.1won 17 · lost 11
Gemini 3.8 Flashwon 18 · lost 12
Kimi K3won 16 · lost 12
GPT-5.5won 17 · lost 13
GPT-5.1won 13 · lost 10
Claude Opus 4.8won 15 · lost 13
Grok 4.6won 16 · lost 14
Claude Opus 4.7won 15 · lost 14
Grok 4.7won 13 · lost 13
GPT-6 Astrawon 8 · lost 16
Claude Opus 5won 10 · lost 25
Claude Fable 5.1won 3 · lost 21
Claude Opus 5.5won 3 · lost 24
Claude Fable 5won 2 · lost 21

§ 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 1483
GDPval-AA ↗1 measureread 2026-09-26
GDPval-AA 1588
AA-Briefcase ↗1 measureread 2026-09-26
AA-Briefcase 1487
Terminal-Bench ↗1 measureread 2026-09-26
Terminal-Bench 37.3% (Codex)
ARC-AGI-2 ↗1 measureread 2026-09-26
ARC-AGI-2 92.5%
LiveBench ↗8 measuresread 2026-09-26
LiveBench 81LiveBench · Reasoning 91.7LiveBench · Coding 83.9LiveBench · Agentic Coding 56.2LiveBench · Mathematics 96.2LiveBench · Data Analysis 79.8LiveBench · Language 87.7LiveBench · Instruction Following 71.8
Kagi LLM Benchmark ↗1 measureread 2026-09-26
Kagi LLM Benchmark 67%
SimpleBench ↗1 measureread 2026-09-26
SimpleBench 64.8%
Vals.ai ↗18 measuresread 2026-09-26
Vals · Legal Research Bench 48.08%Vals · LegalBench 86.97%Vals · Harvey Legal Agent Benchmark 2.5%Vals · Finance Agent 53.76%Vals · CorpFin 64.38%Vals · TaxEval 74.78%Vals · MortgageTax 67.29%Vals · MedCode 43.97%Vals · MedScribe 85.23%Vals · BioMysteryBench 71.11%Vals · SRE Bench 30.53%Vals · Web Search Index 45.24% (Exa)Vals · SWE-bench Verified 96.2% (Mini-SWE-agent)Vals · Vibe Code Bench 80.5% (OpenHands)Vals · Code Migration 52.92%Vals · GPQA Diamond 95.2%Vals · MMLU Pro 89.1%Vals · ProofBench 83%
Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-09-26
Enkrypt · Jailbreak risk 9.2%Enkrypt · Harmful content risk 0.6%Enkrypt · CBRN risk 19.2%Enkrypt · Toxicity risk 8.9%Enkrypt · Bias risk 72.1%Enkrypt · Insecure code risk 7.1%
Vectara · Factual consistency 87.6%
GPQA Diamond (Pass@1) 94.1%HLE (Pass@1) 44.5%Terminal-Bench 2.1 (Pass@1) 88.8%Terminal-Bench 3.0 (Pass@1) 34.4%Terminal-Bench 4.0 (Pass@1) 39.9%DeepSWE v1.1 (Resolved) 73%ProgramBench (Almost@1) 23%NL2Repo-Bench (Score) 56.8%CyberGym (Pass@1) 84.5%SEC-Bench Pro (Pass@1) 74.3%ExploitGym (Pass@1) 33.7%AutomationBench (Pass@1) 45.8%Agent's Last Exam (Pass@1) 26.7%Chartography w/ tools (Pass@1) 79.9%BabyVision w/ tools (Pass@1) 88.9%ZeroBench-main w/ tools (Pass@5) 53%

Badge

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

✱ 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