Inference
EN
OpenAI

GPT-5.6 Sol

GPT-5.6 Sol is a reasoning model from OpenAI in the GPT-5.6 family. 70 benchmarks count toward its index, in 8 categories.

  • Closed weights
  • Reasoning
  • Input: file · image · text
  • Released 2026-07
RankedEnough results for an index.
IndexAll categories, out of 100.73.0±1.9Median 51.0
CoverageIndex weight that has results.100%
SpeedOutput tokens per second.64/sMedian 67/s
First tokenSeconds to the first output token.2.8 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$4.40batch $2.00Median $0.60
Output / 1MUS dollars per 1M output tokens.$22.00batch $10.00 Batch price per 1M output tokens.Median $2.55
ContextMaximum tokens in one request.1.05MMedian 262K
EloLMArena rating and rank.1457±4 #45Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
openai/gpt-5.6-sol
Max outputMaximum output tokens per request.
128K
Knowledge cutoffLast date of training data.
16 Feb 2026
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.44
WeightsLink to the published weights.
—
RetiresDate OpenRouter removes the model.
—

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.21 counted71.916 / 218
CodingCode writing and repair.20 counted70.419 / 209
ReasoningLogic problems and puzzles.10 counted76.312 / 227
MultimodalTasks with images and text.3 counted72.65 / 84
KnowledgeFacts and expert knowledge.8 counted72.711 / 268
MultilingualTasks in many languages.1 counted66.4—
InstructionTasks with strict prompt rules.2 counted64.57 / 63
MathMath problems.5 counted75.114 / 168

Results

70 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
OTIS Mock AIME 2024-2025Math100.0%1 / 200
Critical Physics TasksReasoningAbove fitFar above this model's index.32.3%3 / 214
GPQA Diamond · Vals AIKnowledge95.2% ±1.12 / 135
ARC-AGI-2 (semi-private)Reasoning92.5%3 / 131
Terminal-Bench 2.1 · BenchLMAgentic91.9%2 / 86
Artificial Analysis GPQA DiamondKnowledge94.1%5 / 199
EuroEval SwedishMultilingual74.9% ±0.55 / 195
EuroEval GermanMultilingual63.6% ±0.55 / 191
EuroEval PortugueseMultilingual70.5% ±0.45 / 182
Graduate-Level Google-Proof Q&AKnowledge94.6%4 / 141
BrowseCompAgentic92.2%2 / 63
IOI v1Coding86.7% ±1.72 / 61
EuroEval SpanishMultilingual66.3% ±0.46 / 176
EuroEval DutchMultilingual68.7% ±0.46 / 175
SWE-benchCoding96.2% ±0.93 / 87
Artificial Analysis Long Context ReasoningReasoning84.0%8 / 217
EuroEval FrenchMultilingual72.0% ±0.47 / 178
Terminal-Bench 2.1 · Vals AIAgentic85.8% ±1.43 / 75
EuroEval ItalianMultilingual70.1% ±0.47 / 174
Artificial Analysis ITBench-AAAgentic56.2%1 / 24
Terminal-Bench HardAgentic65.9%1 / 23
GPQA Diamond · BenchLMKnowledge94.6%4 / 84
Artificial Analysis Omniscience AccuracyKnowledge59.4%8 / 164
EuroEval PolishMultilingual62.8% ±0.411 / 209
Artificial Analysis Humanity's Last ExamKnowledge49.5%13 / 222
LiveBench ReasoningReasoning91.7%5 / 80
MMMU ProMultimodal88.8% ±0.86 / 92
LiveBench CodingCoding83.9%5 / 66
ParseBench: document parsing for AI agents (LlamaIndex leaderboard)Multimodal74.4%3 / 39
GPQA diamond · Epoch AIKnowledge93.5% ±1.617 / 217
ARC-AGI-1 (semi-private)Reasoning96.5%11 / 126
Artificial Analysis MMMU-ProMultimodal83.4%10 / 113
Massive Multi-discipline Multimodal Understanding ProMultimodal83.0%5 / 53
FrontierMath-Tiers-1-3-v2-PrivateMath89.1% ±1.89 / 90
MazeBench: visual spatial reasoning in a 3D open worldAgentic13.0%5 / 48
LiveBench MathematicsMath96.2%7 / 66
LiveBench LanguageKnowledge87.7%7 / 66
SimpleQA VerifiedKnowledge69.7% ±1.510 / 91
IOICoding91.2% ±4.55 / 45
MMLU ProKnowledge89.1% ±0.315 / 135
Chess PuzzlesReasoning59.0% ±4.910 / 90
Mystery Game PuzzlesReasoning58.0% ±5.010 / 80
ToolathlonAgentic58.0%3 / 22
FrontierMath-Tier-4-v2-PrivateMath82.9% ±5.910 / 72
GDPval-AA normalizedAgentic55.4%19 / 133
LiveBench Data AnalysisReasoning79.8%12 / 80
Terminal-Bench 3.0Agentic34.6%2 / 13
Code MigrationCoding52.9% ±4.412 / 77
LiveBench Browser UseAgentic88.0%8 / 49
τ²-bench BankingAgentic46.9%5 / 28
ExtractBench: schema-guided enterprise document extraction (LlamaIndex leaderboard)Multimodal93.8%5 / 26
Artificial Analysis SciCodeCoding57.1%22 / 114
SWE-bench ProCoding64.6%18 / 83
Artificial Analysis IFBenchInstruction72.7%34 / 146
LiveBench Long AgenticAgentic48.5%12 / 51
OSWorld 2.0Agentic62.6%6 / 25
Bug Hunt BenchCoding42.0%6 / 25
LiveBench DesignAgentic81.6%13 / 52
Vibe Code Bench v1.1Coding80.5% ±3.731 / 123
DeepSWEAgentic72.7%15 / 59
ApprenticeBench: end-to-end computer use, continual learning, and long-horizon agency on a real accounts-payable jobAgentic26.0%5 / 18
VulcanBench v3Coding87.0%4 / 14
LiveBench Coding 2.0Coding89.7%15 / 52
ProofBench v1.1Math83.0% ±3.816 / 54
Agent Arena steerabilityAgentic5.516 / 54
ExploitGymAgentic33.7%7 / 23
CyberGymAgentic84.5%9 / 29
Terminal-Bench 4.0Agentic37.9%19 / 61
LiveBench Instruction FollowingInstruction71.8%21 / 66
Furniture AssemblyReasoning56.7% ±6.312 / 37
cursorBench32Coding67.2%6 / 18
ARC-AGI-3 (semi-private)Reasoning7.8%9 / 26
EBR-benchReasoning44.8% ±4.211 / 31
LiveCodeBenchCoding82.6% ±1.151 / 142
τ²-Bench Tool-Agent-User EvaluationAgentic85.1%56 / 149
HealthBench HardKnowledge33.1%5 / 13
PostTrainBench v1.1Coding36.2%6 / 15
LiveBench Agentic CodingAgentic56.2%28 / 66
SkillsBenchCoding54.1% ±4.715 / 35
FrontierCode 2.0Coding27.6%12 / 27
Artificial Analysis Tau3-BankingAgentic44.3%10 / 22
Vibe Code Bench 1-100Coding20.0% ±3.612 / 26
Agent Arena task outcomeAgentic2.325 / 54
Terminal-Bench 4.1.0Agentic38.5% ±2.011 / 23
Terminal-Bench 4.0.0Agentic37.3% ±1.911 / 23
FrontierSWE v2Coding32.2%11 / 22
Artificial Analysis AnalystAgentAgentic47.5%14 / 26
HealthBench ProfessionalKnowledge60.5%10 / 18
Agent Arena command recoveryAgentic0.728 / 50
cursorBench40Coding41.7%13 / 21
LiveBench ResearchAgentic54.5%46 / 52
MirrorCodeCoding20.0% ±9.237 / 39
ARC-AGI-1 (public eval)Reasoning99.0%2 / 116
LegalBenchLegal87.0% ±0.47 / 149
ARC-AGI-2 (public eval)Reasoning93.8%10 / 122
CyberBench v1.1Other76.3% ±5.24 / 48
EMBFinance72.3% ±2.47 / 75
Legal Research BenchLegal48.1% ±3.512 / 90
SRE BenchOther30.5% ±2.96 / 36
TaxEval v2Finance74.8% ±0.929 / 143
ProgramBenchCoding1.5% ±0.914 / 65
MortgageTaxFinance67.3% ±0.926 / 97
Terminal-Bench-ScienceScience20.0% ±4.812 / 44
CorpFin v2Finance64.4% ±0.937 / 131
MedScribeHealthcare85.2% ±2.031 / 109
Tax Agent BenchFinance31.1% ±3.329 / 82
Public Benefits Bench v1.1Public benefits66.5% ±1.218 / 50
Time Horizon Index: KSPOther23.8%5 / 13
Finance Agent (v2)Finance53.8% ±0.831 / 79
BioMysteryBenchScience71.1% ±0.612 / 30
OEIS Open LiteMath43.0% ±5.02 / 5
MedCodeHealthcare44.0% ±2.343 / 106
MysteryMechanismScience33.3% ±3.213 / 30
FrontierMath-ErdosMath0.0%8 / 15
CUA-benchOther8.3%5 / 9
Vals Web Search IndexOther43.6% ±2.08 / 12
VulcanBench Coding Intelligence Index v1Coding86.5%—
MMMU-Pro with PythonMultimodal84.6%—
LABBench2: An Improved Benchmark for AI Systems Performing Biology ResearchKnowledge82.1%—
LifeVal: everyday life tasksAgentic80.0%—
FrontierCode 1.1 ExtendedCoding60.6%—
HLE-VerifiedKnowledge54.5%—
BioBasicsBench: everyday wet-lab scienceReasoning45.6%—
Artificial Analysis EnterpriseOps-GymAgentic42.9%—
GeneBench-ProReasoning28.7%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseEpoch AICC BYARC Prize FoundationTerms restrict reuseVals AINo licence statedLiveBenchNo licence statedEuroEvalMITSierra ResearchMITTerminal-BenchNo licence statedLMArenaCC BY 4.0

History

5 releases
Index74.6 → 73.0
23 Oct 202612 Nov 2026
Output / 1M$20.00 → $22.00
23 Oct 202612 Nov 2026

Questions

How much does GPT-5.6 Sol cost?
$4.40 per 1M input tokens, $22.00 per 1M output tokens.
How good is GPT-5.6 Sol?
Its index is 73.0 out of 100. That is place 17 of 283 ranked models.
What is GPT-5.6 Sol best at?
Knowledge: 72.7 out of 100, place 11 of 268.
How fast is GPT-5.6 Sol?
64 output tokens per second, 2.8 s to the first token.
What is the context window of GPT-5.6 Sol?
1.05M tokens.
Is GPT-5.6 Sol open source?
No. Its weights are not published.
When was GPT-5.6 Sol released?
July 2026.

Badge

GPT-5.6 Sol on Inference360: place 17 of 283, index 73.0
HTML
<a href="https://inference360.ai/models/gpt-5-6-sol"><img src="https://inference360.ai/badges/gpt-5-6-sol.svg" alt="GPT-5.6 Sol on Inference360: place 17 of 283, index 73.0" height="20"></a>
Markdown
[![GPT-5.6 Sol on Inference360: place 17 of 283, index 73.0](https://inference360.ai/badges/gpt-5-6-sol.svg)](https://inference360.ai/models/gpt-5-6-sol)

Compare with

Muse Wonder73.1GPT-6 Sol72.8GPT-5.4 Pro73.3Claude Opus 573.8Hy472.0

Lower price

MiMo-V2.6-Pro70.2 · $0.87/1MHy472.0 · $1.25/1MMuse Spark 1.370.1 · $4.25/1M

More from OpenAI

All models
GPT-5.6 Terra68.9GPT-5.6 Luna65.5GPT-5.6 Sol ProUnrankedGPT-5.6 Terra ProUnrankedGPT-5.6 Luna ProUnranked