Inference
EN
OpenAI

GPT-6 Sol

GPT-6 Sol is a reasoning model from OpenAI in the GPT-6 family. 56 benchmarks count toward its index, in 8 categories.

  • Closed weights
  • Reasoning
  • Input: file · image · text
  • Released 2026-09
RankedEnough results for an index.
IndexAll categories, out of 100.72.8±2.2Median 51.0
CoverageIndex weight that has results.100%
SpeedOutput tokens per second.61/sMedian 67/s
First tokenSeconds to the first output token.2.3 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$2.00batch $1.00Median $0.60
Output / 1MUS dollars per 1M output tokens.$10.00batch $5.00 Batch price per 1M output tokens.Median $2.55
ContextMaximum tokens in one request.1.05MMedian 262K
EloLMArena rating and rank.1397±5 #163Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
openai/gpt-6-sol
Max outputMaximum output tokens per request.
128K
Knowledge cutoffLast date of training data.
—
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.20
WeightsLink to the published weights.
—
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
5 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
Amazon BedrockCloud$2.00$10.001.05M
DigitalOceanCloud$2.00$10.001.05M
Microsoft AzureCloud$2.00$10.001.05M
OpenAIMaker$2.00$10.001.05M
OpenRouterRouter$2.00$10.001.05M

US dollars per 1M tokens, list price. From models.dev (opens in a new tab) (MIT).

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.20 counted68.629 / 218
CodingCode writing and repair.8 counted73.914 / 209
ReasoningLogic problems and puzzles.11 counted76.113 / 227
MultimodalTasks with images and text.3 counted70.111 / 84
KnowledgeFacts and expert knowledge.7 counted71.912 / 268
MultilingualTasks in many languages.1 counted65.5—
InstructionTasks with strict prompt rules.1 counted65.3—
MathMath problems.5 counted75.812 / 168

Results

56 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
OTIS Mock AIME 2024-2025Math100.0%1 / 200
EuroEval SwedishMultilingual71.3% ±0.56 / 195
EuroEval GermanMultilingual63.1% ±0.56 / 191
EuroEval PortugueseMultilingual70.1% ±0.46 / 182
Critical Physics TasksReasoningAbove fitFar above this model's index.30.9%8 / 214
EuroEval SpanishMultilingual65.6% ±0.47 / 176
EuroEval DutchMultilingual67.5% ±0.47 / 175
EuroEval ItalianMultilingual69.9% ±0.48 / 174
GPQA diamond · Epoch AIKnowledge94.3% ±1.510 / 217
Artificial Analysis Long Context ReasoningReasoning83.7%10 / 217
Artificial Analysis DeepsecBench-AAAgentic46.3%1 / 21
EuroEval FrenchMultilingual70.4% ±0.49 / 178
EuroEval PolishMultilingual62.4% ±0.414 / 209
ARC-AGI-2 (semi-private)Reasoning89.6%10 / 131
Artificial Analysis Omniscience AccuracyKnowledge54.5%13 / 164
Terminal-Bench 2.1 · Vals AIAgentic83.1% ±1.36 / 75
Artificial Analysis Humanity's Last ExamKnowledge47.9%18 / 222
LiveBench Browser UseAgentic92.0%4 / 49
LiveBench Data AnalysisReasoning81.2%7 / 80
FrontierMath-Tiers-1-3-v2-PrivateMath89.8% ±1.88 / 90
LiveBench MathematicsMath96.4%6 / 66
Vibe Code Bench v1.1Coding87.8% ±2.512 / 123
Artificial Analysis MMMU-ProMultimodal82.9%12 / 113
FrontierMath-Tier-4-v2-PrivateMath90.0% ±5.28 / 72
ExtractBench: schema-guided enterprise document extraction (LlamaIndex leaderboard)Multimodal94.6%3 / 26
Code MigrationCoding57.2% ±4.29 / 77
LiveBench Long AgenticAgentic66.9%6 / 51
ARC-AGI-1 (semi-private)Reasoning95.5%17 / 126
Agents' Last ExamAgentic56.4%3 / 21
Mystery Game PuzzlesReasoning56.0% ±5.012 / 80
Artificial Analysis CWE-Bench-AAAgentic64.2%3 / 19
Artificial Analysis SciCodeCoding57.6%18 / 114
LiveBench CodingCoding81.8%11 / 66
Chess PuzzlesReasoning56.0% ±5.015 / 90
LiveBench Coding 2.0Coding93.1%9 / 52
LiveBench DesignAgentic82.2%9 / 52
SimpleQA VerifiedKnowledge60.7% ±1.517 / 91
LiveBench LanguageKnowledge85.3%13 / 66
GDPval-AA normalizedAgentic50.4%27 / 133
Agent Arena steerabilityAgentic10.111 / 54
Terminal-Bench 4.0Agentic44.4% ±3.513 / 61
Agent Arena command recoveryAgentic7.011 / 50
IOICoding82.6% ±9.210 / 45
LiveBench ReasoningReasoning88.7%20 / 80
MazeBench: visual spatial reasoning in a 3D open worldAgentic7.0%12 / 48
FrontierCode 2.0Coding34.3%7 / 27
EBR-benchReasoning53.3% ±4.69 / 31
Artificial Analysis ITBench-AAAgentic49.4%7 / 24
ProofBench v1.1Math83.0% ±3.816 / 54
Furniture AssemblyReasoning58.3% ±6.411 / 37
Terminal-Bench 4.1.0Agentic50.3% ±1.67 / 23
Terminal-Bench 4.0.0Agentic49.4% ±1.67 / 23
Hermes Index: agentic capability and cost per task inside Hermes AgentAgentic44.1%5 / 16
Agent Arena task outcomeAgentic4.517 / 54
Medical Long Context Reasoning (MLCR-AA)Reasoning33.9%7 / 22
OSWorld 2.0Agentic60.5%8 / 25
ARC-AGI-3 (semi-private)Reasoning4.6%10 / 26
DeepSWEAgentic68.8%24 / 59
ParseBench: document parsing for AI agents (LlamaIndex leaderboard)Multimodal68.8%16 / 39
HealthBench ProfessionalKnowledge60.8%8 / 18
MirrorCodeCoding48.9% ±11.218 / 39
LiveBench Instruction FollowingInstruction68.6%33 / 66
ExploitGymAgentic22.1%12 / 23
Artificial Analysis Cyber IndexAgentic36.8%10 / 19
LiveBench Agentic CodingAgentic52.9%36 / 66
LiveBench ResearchAgentic69.3%34 / 52
AutomationBenchAgentic33.2%23 / 32
HealthBench HardKnowledge30.1%10 / 13
Artificial Analysis CyberGym-E2E-AAAgentic0.0%17 / 19
CWE-bench v1Agentic52.0%11 / 12
CyberBench v1.1Other78.0% ±5.12 / 48
ARC-AGI-1 (public eval)Reasoning97.8%14 / 116
ARC-AGI-2 (public eval)Reasoning90.0%17 / 122
EMBFinance71.5% ±2.412 / 75
SRE BenchOther30.5% ±2.96 / 36
ProgramBenchCoding2.0% ±1.011 / 65
Terminal-Bench-ScienceScience30.0% ±5.59 / 44
BioMysteryBenchScience74.8% ±2.48 / 30
MedCodeHealthcare47.1% ±2.136 / 106
FrontierMath-ErdosMath1.5% ±1.56 / 15
MedScribeHealthcare82.0% ±1.952 / 109
MysteryMechanismScience30.2% ±3.115 / 30
SpireClimb (Slay the Spire 2, Ascension 10)Other13.3% ±4.05 / 9
SpireClimbOther9.2% ±2.45 / 9
SpireClimb (Slay the Spire 1, Ascension 20 Heart)Other5.0% ±2.85 / 9
Finance Agent (v2)Finance49.0% ±0.648 / 79
Legal Research BenchLegal28.8% ±3.157 / 90
Tax Agent BenchFinance15.0% ±2.662 / 82
Public Benefits Bench v1.1Public benefits56.6% ±1.339 / 50
Hermes Bench: 150 agent tasks graded on the files and state the agent leaves behindAgentic73.5%—
HealthBench Professional raw scoreKnowledge59.5%—
HealthBench length-adjusted scoreKnowledge53.2%—
HealthBench raw scoreKnowledge47.1%—
Harvey LAB-AA v1.1Agentic3.6%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseEpoch AICC BYLiveBenchNo licence statedARC Prize FoundationTerms restrict reuseVals AINo licence statedEuroEvalMITTerminal-BenchNo licence statedLMArenaCC BY 4.0models.devMIT

History

5 releases
Index75.0 → 72.8
23 Oct 202612 Nov 2026
Output / 1M$10.00 → $10.00
23 Oct 202612 Nov 2026

Questions

How much does GPT-6 Sol cost?
$2.00 per 1M input tokens, $10.00 per 1M output tokens.
How good is GPT-6 Sol?
Its index is 72.8 out of 100. That is place 18 of 283 ranked models.
What is GPT-6 Sol best at?
Knowledge: 71.9 out of 100, place 12 of 268.
How fast is GPT-6 Sol?
61 output tokens per second, 2.3 s to the first token.
What is the context window of GPT-6 Sol?
1.05M tokens.
Is GPT-6 Sol open source?
No. Its weights are not published.
When was GPT-6 Sol released?
September 2026.

Badge

GPT-6 Sol on Inference360: place 18 of 283, index 72.8
HTML
<a href="https://inference360.ai/models/gpt-6-sol"><img src="https://inference360.ai/badges/gpt-6-sol.svg" alt="GPT-6 Sol on Inference360: place 18 of 283, index 72.8" height="20"></a>
Markdown
[![GPT-6 Sol on Inference360: place 18 of 283, index 72.8](https://inference360.ai/badges/gpt-6-sol.svg)](https://inference360.ai/models/gpt-6-sol)

Compare with

GPT-5.6 Sol73.0Muse Wonder73.1GPT-5.4 Pro73.3Hy472.0Claude Opus 573.8

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-6 Luna65.5GPT-6 Sol ProUnrankedGPT-6 Luna ProUnrankedGPT-6.1 Sol78.0GPT-6.1 Sol ProUnranked