Inference
EN
OpenAI

GPT-5.6 Terra

GPT-5.6 Terra is a reasoning model from OpenAI in the GPT-5.6 family. 66 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 100.68.9±2.0Median 51.0
CoverageIndex weight that has results.100%
SpeedOutput tokens per second.66/sMedian 67/s
First tokenSeconds to the first output token.2.1 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$2.00batch $1.00Median $0.60
Output / 1MUS dollars per 1M output tokens.$12.00batch $6.00 Batch price per 1M output tokens.Median $2.55
ContextMaximum tokens in one request.1.05MMedian 262K
EloLMArena rating and rank.1449±4 #55Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
openai/gpt-5.6-terra
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.
$0.20
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.22 counted68.233 / 218
CodingCode writing and repair.16 counted68.424 / 209
ReasoningLogic problems and puzzles.9 counted72.520 / 227
MultimodalTasks with images and text.3 counted67.316 / 84
KnowledgeFacts and expert knowledge.8 counted65.542 / 268
MultilingualTasks in many languages.1 counted60.7—
InstructionTasks with strict prompt rules.2 counted56.421 / 63
MathMath problems.5 counted72.818 / 168

Results

66 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
EuroEval SwedishMultilingual70.6% ±0.59 / 195
Critical Physics TasksReasoningAbove fitFar above this model's index.30.0%10 / 214
EuroEval GermanMultilingual59.7% ±0.610 / 191
EuroEval FrenchMultilingual68.9% ±0.410 / 178
SWE-benchCoding95.4% ±0.95 / 87
EuroEval ItalianMultilingual67.8% ±0.410 / 174
EuroEval DutchMultilingual65.0% ±0.511 / 175
EuroEval SpanishMultilingual62.1% ±0.412 / 176
OTIS Mock AIME 2024-2025Math99.7% ±0.314 / 200
Artificial Analysis Long Context ReasoningReasoning83.0%18 / 217
Graduate-Level Google-Proof Q&AKnowledge92.9%12 / 141
ARC-AGI-1 (semi-private)Reasoning96.5%11 / 126
GPQA diamond · Epoch AIKnowledge93.3% ±1.519 / 217
EuroEval PortugueseMultilingual65.4% ±0.516 / 182
Artificial Analysis GPQA DiamondKnowledge92.5%18 / 199
IOI v1Coding65.3% ±0.97 / 61
EuroEval PolishMultilingual59.0% ±0.524 / 209
ARC-AGI-2 (semi-private)Reasoning83.9%17 / 131
APEX-Agents-AAAgentic38.9%4 / 30
LiveBench ReasoningReasoning90.6%11 / 80
FrontierMath-Tiers-1-3-v2-PrivateMath86.0% ±2.113 / 90
Artificial Analysis MMMU-ProMultimodal80.7%17 / 113
Artificial Analysis Humanity's Last ExamKnowledge42.9%34 / 222
GPQA Diamond · BenchLMKnowledge92.9%13 / 84
IOICoding87.6% ±6.57 / 45
Terminal-Bench 2.1 · BenchLMAgentic87.4%14 / 86
Code MigrationCoding47.8% ±4.313 / 77
LiveCodeBenchCoding85.9% ±1.025 / 142
GPQA Diamond · Vals AIKnowledge90.9% ±1.924 / 135
Terminal-Bench 2.1 · Vals AIAgentic77.5% ±2.214 / 75
LiveBench Data AnalysisReasoning79.3%15 / 80
Artificial Analysis Omniscience AccuracyKnowledge46.8%31 / 164
Artificial Analysis ITBench-AAAgentic51.0%5 / 24
MMMU ProMultimodal86.5% ±0.820 / 92
BrowseCompAgentic87.5%14 / 63
LiveBench Browser UseAgentic84.0%11 / 49
LiveBench MathematicsMath94.9%15 / 66
FrontierMath-Tier-4-v2-PrivateMath70.7% ±7.217 / 72
Artificial Analysis Terminal-Bench v2.1Agentic88.0%5 / 21
SWE-bench ProCoding63.4%20 / 83
Massive Multi-discipline Multimodal Understanding ProMultimodal80.7%13 / 53
GDPval-AA normalizedAgentic47.6%33 / 133
Artificial Analysis SciCodeCoding55.0%29 / 114
LiveBench Long AgenticAgentic48.4%13 / 51
Chess PuzzlesReasoning47.0% ±5.024 / 90
Artificial Analysis IFBenchInstruction71.2%39 / 146
ExtractBench: schema-guided enterprise document extraction (LlamaIndex leaderboard)Multimodal92.3%7 / 26
MazeBench: visual spatial reasoning in a 3D open worldAgentic6.0%13 / 48
VulcanBench v3Coding87.0%4 / 14
Mystery Game PuzzlesReasoning35.0% ±4.823 / 80
LiveBench LanguageKnowledge82.9%19 / 66
MMLU ProKnowledge86.7% ±0.339 / 135
SkillsBenchCoding58.9% ±4.511 / 35
τ²-Bench Tool-Agent-User EvaluationAgentic86.3%51 / 149
Furniture AssemblyReasoning54.2% ±6.413 / 37
Vibe Code Bench v1.1Coding74.6% ±4.244 / 123
ToolathlonAgentic53.1%8 / 22
ProofBench v1.1Math74.0% ±4.420 / 54
LiveBench Coding 2.0Coding86.2%20 / 52
DeepSWEAgentic69.6%23 / 59
OSWorld 2.0Agentic50.2%10 / 25
Artificial Analysis DeepsecBench-AAAgentic25.1%9 / 21
LiveBench CodingCoding78.2%29 / 66
LiveBench Agentic CodingAgentic54.9%29 / 66
cursorBench32Coding64.9%8 / 18
ParseBench: document parsing for AI agents (LlamaIndex leaderboard)Multimodal67.1%18 / 39
HealthBench HardKnowledge32.7%6 / 13
Terminal-Bench 3.0Agentic20.8%6 / 13
ExploitGymAgentic23.2%11 / 23
LiveBench DesignAgentic76.0%26 / 52
Agent Arena steerabilityAgentic−0.527 / 54
SimpleQA VerifiedKnowledge43.2% ±1.646 / 91
Terminal-Bench 4.0Agentic22.7% ±2.332 / 61
MirrorCodeCoding41.1% ±12.421 / 39
Agent Arena command recoveryAgentic0.927 / 50
CyberGymAgentic81.8%16 / 29
ApprenticeBench: end-to-end computer use, continual learning, and long-horizon agency on a real accounts-payable jobAgentic16.0%10 / 18
ARC-AGI-3 (semi-private)Reasoning0.8%15 / 26
Artificial Analysis Tau3-BankingAgentic40.2%14 / 22
LiveBench Instruction FollowingInstruction64.6%43 / 66
Terminal-Bench 4.0.0Agentic21.5% ±1.715 / 23
Terminal-Bench 4.1.0Agentic21.5% ±1.715 / 23
Vibe Code Bench 1-100Coding14.8% ±3.119 / 26
Agent Arena task outcomeAgentic−7.241 / 54
cursorBench40Coding41.3%16 / 21
HealthBench ProfessionalKnowledge57.7%14 / 18
LiveBench ResearchAgentic60.4%43 / 52
TaxEval v2Finance76.2% ±0.86 / 143
ARC-AGI-1 (public eval)Reasoning98.3%8 / 116
ARC-AGI-2 (public eval)Reasoning91.3%13 / 122
LegalBenchLegal85.1% ±0.525 / 149
CorpFin v2Finance65.3% ±0.928 / 131
MortgageTaxFinance67.3% ±0.926 / 97
ProgramBenchCoding0.5% ±0.519 / 65
EMBFinance66.2% ±3.123 / 75
Legal Research BenchLegal41.3% ±3.430 / 90
Finance Agent (v2)Finance54.4% ±2.127 / 79
Terminal-Bench-ScienceScience10.0% ±3.616 / 44
CyberBench v1.1Other72.1% ±5.418 / 48
SAGEEducation87.0% ±2.011 / 29
SRE BenchOther9.5% ±1.814 / 36
Tax Agent BenchFinance30.0% ±3.334 / 82
MedCodeHealthcare43.4% ±2.245 / 106
MedScribeHealthcare82.9% ±1.950 / 109
Public Benefits Bench v1.1Public benefits62.4% ±1.327 / 50
MMMU-Pro with PythonMultimodal82.0%—
LABBench2: An Improved Benchmark for AI Systems Performing Biology ResearchKnowledge81.2%—
FrontierCode 1.1 ExtendedCoding55.8%—
HLE-VerifiedKnowledge51.1%—

Sources

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

History

5 releases
Index70.2 → 68.9
23 Oct 202612 Nov 2026
Output / 1M$12.00 → $12.00
23 Oct 202612 Nov 2026

Questions

How much does GPT-5.6 Terra cost?
$2.00 per 1M input tokens, $12.00 per 1M output tokens.
How good is GPT-5.6 Terra?
Its index is 68.9 out of 100. That is place 26 of 283 ranked models.
What is GPT-5.6 Terra best at?
Reasoning: 72.5 out of 100, place 20 of 227.
How fast is GPT-5.6 Terra?
66 output tokens per second, 2.1 s to the first token.
What is the context window of GPT-5.6 Terra?
1.05M tokens.
Is GPT-5.6 Terra open source?
No. Its weights are not published.
When was GPT-5.6 Terra released?
July 2026.

Badge

GPT-5.6 Terra on Inference360: place 26 of 283, index 68.9
HTML
<a href="https://inference360.ai/models/gpt-5-6-terra"><img src="https://inference360.ai/badges/gpt-5-6-terra.svg" alt="GPT-5.6 Terra on Inference360: place 26 of 283, index 68.9" height="20"></a>
Markdown
[![GPT-5.6 Terra on Inference360: place 26 of 283, index 68.9](https://inference360.ai/badges/gpt-5-6-terra.svg)](https://inference360.ai/models/gpt-5-6-terra)

Compare with

GPT-5.569.1Grok 4.769.1DeepSeek V4.1 Flash 101768.5Gemini 3.7 Flash68.5MiniMax M3.1 Flash69.3

Lower price

Ling 3.1 Flash67.2 · $0.135/1MDeepSeek V4.1 Flash 101768.5 · $0.40/1MClaude Haiku 5.566.3 · $0.50/1M

More from OpenAI

All models
GPT-5.6 Sol73.0GPT-5.6 Luna65.5GPT-5.6 Sol ProUnrankedGPT-5.6 Terra ProUnrankedGPT-5.6 Luna ProUnranked