BenchLeader

GPT-5.2 vs GPT-5.6 Terra

Full profiles: GPT-5.2 (OpenAI) and GPT-5.6 Terra (OpenAI): every benchmark, price and rank.

Updated · Data from 11 sources

As of 16 Nov 2026, GPT-5.2 and GPT-5.6 Terra are neck and neck on quality (59.9 vs 60.1 on the BenchLeader Index, inside the ±5.7 margin), and they cost about the same at $4.81 vs $4.50 per 1M tokens blended. Choose GPT-5.2 for instruction following (74 vs 66), reasoning (60 vs 55), and coding (63 vs 59); choose GPT-5.6 Terra for fast first answers (19.2 s vs 112.5 s) and agents & tools (63 vs 54) and maths (66 vs 57), where it also streams faster.

Key facts (as of 16 Nov 2026)

  • GPT-5.2 scores 59.9 on the BenchLeader Index and GPT-5.6 Terra scores 60.1 (BenchLeader, 16 Nov 2026).
  • GPT-5.2 costs $4.81 per 1M tokens blended (list price, 16 Nov 2026).
  • GPT-5.6 Terra costs $4.50 per 1M tokens blended (list price, 16 Nov 2026).
  • GPT-5.6 Terra streams faster than GPT-5.2 (Artificial Analysis, 7-day median, 16 Nov 2026).
  • Of the 21 benchmarks both models have, GPT-5.6 Terra leads 14 (BenchLeader, 16 Nov 2026).

Which to pick for…

The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.

Use casePickWhyConfidence
CodingGPT-5.263 vs 59 category score; LiveCodeBench 85.4% vs 85.9%Medium
Agents & tool useGPT-5.6 Terra63 vs 54 category score; Terminal-Bench Hard 62.9% vs 47.0%High
MathsGPT-5.6 Terra66 vs 57 category score; ProofBench 74.0% vs 15.0%High
ReasoningGPT-5.260 vs 55 category score; ARC-AGI-2 52.9% vs 74.2%Medium
KnowledgeGPT-5.256 vs 53 category score; CorpFin 65.9% vs 65.3%Medium
Long documentsGPT-5.266 vs 64 category score; AA-LCR v1.1 82.7% vs 79.0%Medium
Following instructionsGPT-5.274 vs 66 category score; IFBench 75.4% vs 66.3%High
Chat quality (human preference)–Only GPT-5.6 Terra has an Arena rating (66 category score)Not comparable
Lowest costEither$4.81 vs $4.50 per 1M blended; agentic coding session $0.161 vs $0.168–
Fastest first answerGPT-5.6 Terra19.2 s (GPT-5.6 Terra (xhigh)) vs 112.5 s (GPT-5.2 (xhigh)) to the first answer token–
High volume on a budgetNeither flagshipA smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs–

Key differences

  • GPT-5.2 (xhigh) and GPT-5.6 Terra (xhigh) are neck and neck on quality (59.9 vs 60.1, within the ±5.7 margin).
  • GPT-5.2 (xhigh) is stronger in coding, instruction following, knowledge, long context, reasoning.
  • GPT-5.6 Terra (xhigh) is stronger in agents & tools, maths.
  • They cost about the same ($4.50 per 1M blended).

At a glance

SpecGPT-5.2GPT-5.6 Terra
DeveloperOpenAIOpenAI
Released11 Dec 20259 Jul 2026
Context window (tokens)400,0001,050,000
Compared atGPT-5.2 (xhigh)GPT-5.6 Terra (xhigh)
Open weightsNoNo
Price in / out / cached input per 1M$1.75 / $14.00 / $0.175$2.00 / $12.00 / $0.200

Category by category

Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.

CategoryGPT-5.2 (xhigh)GPT-5.6 Terra (xhigh)Leader
Reasoning60.454.8GPT-5.2
Coding63.059.1GPT-5.2
Agents & tools53.863.1GPT-5.6 Terra
Maths56.565.7GPT-5.6 Terra
Knowledge56.153.0GPT-5.2
Instruction following73.765.6GPT-5.2
Human preference–65.5not comparable
Multimodal–61.2not comparable
Long context66.064.1GPT-5.2

On one chart

Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.

2550 avg75ReasoningCodingAgents & toolsMathsKnowledgeInstructionfollowingLong contextGPT-5.2, Reasoning: 60.4GPT-5.2, Coding: 63.0GPT-5.2, Agents & tools: 53.8GPT-5.2, Maths: 56.5GPT-5.2, Knowledge: 56.1GPT-5.2, Instruction following: 73.7GPT-5.2, Long context: 66.0GPT-5.6 Terra, Reasoning: 54.8GPT-5.6 Terra, Coding: 59.1GPT-5.6 Terra, Agents & tools: 63.1GPT-5.6 Terra, Maths: 65.7GPT-5.6 Terra, Knowledge: 53.0GPT-5.6 Terra, Instruction following: 65.6GPT-5.6 Terra, Long context: 64.1

Not drawn, measured for fewer than two of these models: Multimodal, Human preference.

Model, click to add or removeIndex

Where sources disagree

Across the 21 benchmarks both have, GPT-5.2 leads 7, GPT-5.6 Terra leads 14. Every overall index that rates both puts GPT-5.6 Terra ahead.

Overall indexGPT-5.2 (xhigh)GPT-5.6 Terra (xhigh)Leader
BenchLeader Index59.960.1neck and neck
Arena Text1437†1468GPT-5.6 Terra

Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).

Price, speed and latency

List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.

MeasureGPT-5.2 (xhigh)GPT-5.6 Terra (xhigh)
Blended price per 1M tokens$4.81$4.50
Input / output per 1M$1.75 / $14.00$2.00 / $12.00
Cached input per 1M$0.175$0.200
Price per task: chat reply$0.0049$0.0044
Price per task: summarize a 30-page report$0.029$0.031
Price per task: code edit$0.032$0.030
Price per task: agentic coding session$0.161$0.168
Price per task: structured extraction$0.0063$0.0064
Agentic coding session, input mostly cached$0.090$0.087
Output speed76 tok/s97 tok/s
Time to first answer112.5 s19.2 s
Time to a complete answer, last measured 13 Nov 2026154.4 s84.0 s

Each model is compared at its best-ranked reasoning-effort setting. GPT-5.2 (xhigh) scores 59.9 and takes 112 s to its first answer token; GPT-5.2 scores 56.7 in 2 s, GPT-5.2 (medium) scores 56.2 in 7 s. GPT-5.6 Terra (xhigh) scores 60.1 and takes 19 s to its first answer token; GPT-5.6 Terra (max) scores 59.5 in 73 s, GPT-5.6 Terra (high) scores 56.8 in 6 s. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.

Every benchmark

182 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.

MetricPublisherGPT-5.2 (xhigh)GPT-5.6 Terra (xhigh)
BenchLeader IndexBenchLeader59.960.1
Agents & tools scoreBenchLeader53.863.1
Coding scoreBenchLeader63.059.1
Instruction following scoreBenchLeader73.765.6
Knowledge scoreBenchLeader56.153.0
Long context scoreBenchLeader66.064.1
Maths scoreBenchLeader56.565.7
Reasoning scoreBenchLeader60.454.8
Human preference scoreBenchLeader–65.5
Multimodal scoreBenchLeader–61.2
Blended price $/MBenchLeader$4.81$4.50
Output speedBenchLeader76 tok/s97 tok/s
Time to first answerBenchLeader112.5 s19.2 s
Context windowBenchLeader400,000 tokens1,050,000 tokens
Reasoning
GPQA DiamondEpoch AI Benchmarking Hub91.4%93.3%†
Humanity's Last ExamScale AI / CAIS27.8%†–
SimpleBenchSimpleBench45.8%†48.9%
Arena Hard PromptsArena (LMArena)1497†1491
LiveBench 2.0 ReasoningLiveBench–90.6%†
LiveBench 1.0 ReasoningLiveBench83.2%†90.6%†
GPQA Diamond (AA)Artificial Analysis90.3%90.8%
Humanity's Last Exam (AA)Artificial Analysis37.7%41.9%
HLE-Diamond (AA)Artificial Analysis–24.7%
GPQA Diamond (Vals)Vals AI91.7%90.9%
Kagi LLM BenchmarkKagi LLM Benchmark73.3%†51.3%†
ARC-AGI-1ARC Prize86.2%94.0%
ARC-AGI-2ARC Prize52.9%74.2%
ARC-AGI-3ARC Prize–0.7%
CritPtArtificial Analysis11.6%27.1%
CritPt v2 hard subset (AA)Artificial Analysis–50.5%†
SciPredictScale AI SEAL20.6%†–
Arena ExpertArena (LMArena)1491†1507
Chess PuzzlesEpoch AI Benchmarking Hub48.0%47.0%†
Mystery Game PuzzlesEpoch AI Benchmarking Hub22.0%†35.0%†
CL-benchEpoch AI Benchmarking Hub18.2%†–
LMCAEpoch AI Benchmarking Hub43.9%52.2%
DTBenchEpoch AI Benchmarking Hub90.9%92.5%
ForecastBenchForecastBench63.8%†–
Coding
SWE-bench Verified (Epoch)Epoch AI Benchmarking Hub73.8%†–
Show all 182 benchmarks
MetricPublisherGPT-5.2 (xhigh)GPT-5.6 Terra (xhigh)
Coding
SciCodeSciCode–52.3%
WeirdMLWeirdML72.2%78.3%†
FrontierCodeCognition–41.3%†
GSO-BenchGSO-Bench27.4%†–
Arena CodingArena (LMArena)1515†1517
Arena WebDevArena (LMArena)1421†1520
LiveBench 2.0 CodingLiveBench–84.2%†
LiveBench 1.0 CodingLiveBench76.1%†78.3%†
LiveCodeBenchVals AI85.4%85.9%
IOIVals AI–87.6%†
SWE-bench (Vals)Vals AI75.8%95.4%†
SWE-Bench ProScale AI SEAL29.9%†–
SWE-bench Verified (bash only)SWE-bench69.0%†–
SWE-bench Multilingual (bash only)SWE-bench66.7%†–
SWE-bench Verified (any scaffold)SWE-bench72.8%†–
Code MigrationVals AI–47.8%†
ProgramBenchVals AI–0.5%†
Vibe Code Bench 1-100Vals AI–14.8%†
Vibe Code Bench v1.1Vals AI53.5%74.6%†
SWE-Bench Pro (private)Scale AI SEAL23.8%†–
SWE-Bench Pro V2Scale AI SEAL–92.4%
SWE-Bench Pro V2 HARDScale AI SEAL–86.3%
MirrorCodeEpoch AI Benchmarking Hub32.2%41.1%†
Surface Evolver BenchEpoch AI Benchmarking Hub–83.8%
DeepSWE v1.1Epoch AI Benchmarking Hub–60.2%
CursorBenchCursor–33.6%
ALE-BenchEpoch AI Benchmarking Hub1249.8†1951.4†
AlgoTuneAlgoTune2.0†–
WeirdML v3WeirdML5.0%–
Terminal-Bench 4.0 (AA)Artificial Analysis–10.1%
Terminal-Bench 2.1 (AA)Artificial Analysis–80.2%
Startrise FrontendStartrise–66.9%†
GameBench 2 OverallDuelLab GameBench 2–45.6†
GameBench 2 by reasoning settingDuelLab GameBench 2–53.5
Design Arena WebsiteDesign Arena11861260†
FrontierCode 2.0 (AA)Artificial Analysis–23.9%†
LiveCodeBench (AA)Artificial Analysis88.9%–
Agents & tools
Terminal-BenchTerminal-Bench64.9%†21.5%†
GDPvalOpenAI49.7%†–
Remote Labor IndexScale AI / CAIS2.1%†–
APEX-AgentsMercor–58.2%†
Arena AgentArena (LMArena)–-1.8
LiveBench Long AgenticLiveBench–25.9%†
LiveBench BrowserLiveBench–70.3%†
LiveBench Agentic CodingLiveBench50.3%†55.0%†
Terminal-Bench HardArtificial Analysis47.0%62.9%
τ²-Bench Telecom (AA)Artificial Analysis84.8%80.4%
Terminal-Bench 2.1 (Vals)Vals AI–77.5%†
MCP AtlasScale AI SEAL67.6%–
BFCL OverallBerkeley Function Calling Leaderboard55.9%†–
τ²-benchτ²-bench84.8%†–
τ³-bench Bankingτ²-bench32.2%†–
GDPval-AA v2.1, normalised scoreArtificial Analysis–47.4%
GDPval-AA v2.1Artificial Analysis–1456
τ³-Banking (AA)Artificial Analysis11.1%†29.7%
ITBench SRE (AA)Artificial Analysis–51.0%†
APEX-Agents (AA)Artificial Analysis–38.9%†
CyberBenchVals AI–72.1%†
Finance Agent v2Vals AI–54.4%†
Harvey's Legal Agent Benchmark v1Vals AI–0.8%†
Legal Research BenchVals AI–41.4%†
Poker AgentVals AI1131.8%†–
SkillsBenchVals AI–58.9%†
SREBenchVals AI–9.5%†
Tax Agent BenchVals AI–30.0%†
Terminal-Bench 2.0 (Vals)Vals AI51.7%†–
Terminal-Bench 4.0 (Vals)Vals AI–22.7%†
Terminal-Bench ScienceVals AI–10.0%†
Arena SearchArena (LMArena)1161†–
Arena Agent Task CompletionArena (LMArena)–-7.2
BALROGBALROG–53.2%†
DeepResearch BenchDeepResearch Bench41.1%†–
METR Time HorizonsMETR75.3%†–
Vending-Bench 2Andon Labs3591.3†7343.2†
AutomationBenchArtificial Analysis–47.1%
GDP.pdf (AA)Artificial Analysis–24.6%
EnterpriseOps-GymArtificial Analysis–38.5%†
AA-Briefcase v1.1Artificial Analysis–1332
Design Arena AgenticDesign Arena1065†1089†
Terminal-Bench-Science 0.1 (AA)Artificial Analysis–13.8%†
DeepSec-Bench (AA)Artificial Analysis–25.1%†
FACTS SearchGoogle–87.0%†
Maths
FrontierMath Tiers 1–3Epoch AI Benchmarking Hub67.4%86.0%†
FrontierMath Tier 4Epoch AI Benchmarking Hub31.7%70.7%†
OTIS Mock AIMEEpoch AI Benchmarking Hub96.1%99.7%†
ProofBenchVals AI15.0%74.0%
LiveBench MathematicsLiveBench93.2%†94.9%†
AIME (Vals)Vals AI96.9%–
MGSMVals AI94.0%–
AIME 2026MathArena98.3%†–
HMMT February 2026MathArena97.0%†–
MathArena ApexMathArena13.5%†–
Arena MathsArena (LMArena)1453†1478
AIME 2025 (AA)Artificial Analysis99.0%–
Knowledge
SimpleQA VerifiedEpoch AI Benchmarking Hub37.1%43.2%†
LiveBench 2.0 Data AnalysisLiveBench–79.3%†
LiveBench ResearchLiveBench–79.7%†
LiveBench 1.0 Data AnalysisLiveBench78.2%†79.3%†
AA-Omniscience v1.1Artificial Analysis–-11.5
MMLU-ProVals AI86.2%86.7%
LegalBenchVals AI82.8%85.1%†
CorpFinVals AI65.9%65.3%
TaxEvalVals AI75.8%76.2%
MedQAVals AI94.1%–
MultiNRCScale AI SEAL42.2%†–
CaseLaw v2Vals AI66.0%–
Excel Modeling BenchmarkVals AI–66.2%†
MedCodeVals AI49.8%43.4%
MedScribeVals AI84.4%82.9%
Public Benefits Bench v1.1Vals AI–62.4%†
Financial Coaching BenchVals AI–71.0%†
DrugDiscoveryBenchScale AI SEAL29.3%–
EBR-benchEpoch AI Benchmarking Hub23.0%–
GDP.pdfEpoch AI Benchmarking Hub–24.7%†
AA-Omniscience v1.1: accuracyArtificial Analysis–41.3%
AA-Omniscience v1.1: non-hallucinationArtificial Analysis–9.9%
Global-MMLU-Lite (AA)Artificial Analysis89.8%†–
FACTS ParametricGoogle–57.6%†
Instruction following
LiveBench LanguageLiveBench79.8%†82.9%†
LiveBench Instruction FollowingLiveBench61.8%†64.6%†
IFBenchArtificial Analysis75.4%66.3%
TutorBenchScale AI SEAL53.5%†–
Arena Instruction FollowingArena (LMArena)1457†1463
Human preference
Arena TextArena (LMArena)1476†1468
EQ-Bench 4EQ-Bench–1220†
EQ-Bench Creative Writing v3EQ-Bench1703†1856†
Arena Creative WritingArena (LMArena)1432†1430
Arena Multi-turnArena (LMArena)1495†1475
Arena Longer QueriesArena (LMArena)1472†1469
Multimodal
Arena VisionArena (LMArena)1268†1269
LiveBench DesignLiveBench–70.9%†
MMMU-ProArtificial Analysis74.6%†79.5%
VISTAScale AI SEAL46.6%†–
MMMU-Pro (Vals)Vals AI86.7%86.5%†
MortgageTaxVals AI67.1%67.3%
SAGEVals AI–87.0%†
Humanity's Sixth SenseScale AI SEAL–22.7%†
VPCTEpoch AI Benchmarking Hub84.0%–
Blueprint-Bench 2Epoch AI Benchmarking Hub–30.8%†
Furniture AssemblyEpoch AI Benchmarking Hub38.3%54.2%†
FACTS MultimodalGoogle–39.9%†
Long context
AA-LCR v1.1Artificial Analysis82.7%79.0%
Arena DocumentArena (LMArena)1425†1474
MLCRArtificial Analysis–22.8%
FACTS GroundingGoogle–60.5%†
Composite
Epoch Capabilities IndexEpoch AI Benchmarking Hub154.5†160.7†
LiveBench 2.0LiveBench–67.2%†
LiveBench 1.0LiveBench74.6%†77.9%†
AA Intelligence Index v5.0.1Artificial Analysis–32.3†
AA Intelligence Index v4.3.2Artificial Analysis30.438.0
AA Progress Index v1Artificial Analysis295.5312
Vals RSI IndexVals AI15.8%–
Vals Multimodal IndexVals AI–65.1%
Safety & honesty
FORTRESSScale AI SEAL17.5%†–
MASKScale AI SEAL86.7%†–
PropensityBenchScale AI SEAL34.4%†–
DistressBenchScale AI SEAL–76.4%†

Data as of 2026-11-16. Best configuration of each model; every score links to its source on the model pages.

DuelLab GameBench 2, release gb2-20261111-7b598bb13de4 (https://benchmarks.duellab.org/releases/gb2-20261111-7b598bb13de4/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.

What changed

  • GPT-5.2 (xhigh): 1 new result (AIME 2025 (AA)).
  • GPT-5.6 Terra (xhigh): 2 new results (GDPval-AA v2.1, AA-Briefcase v1.1).
  • GPT-5.2 (xhigh): 1 new result (MirrorCode).
  • GPT-5.6 Terra (xhigh): 1 score revision.
  • GPT-5.6 Terra (xhigh): 1 score revision.
  • GPT-5.6 Terra (xhigh): BenchLeader Index 61.2 → 60.4.
  • GPT-5.2 (xhigh): 1 new result (AA Progress Index v1).
  • GPT-5.6 Terra (xhigh): 2 new results (AA Progress Index v1, HLE-Diamond (AA)); 3 score revisions.
  • GPT-5.2 (xhigh): 1 new result (Vals RSI Index).
  • GPT-5.6 Terra (xhigh): BenchLeader Index 61.7 → 61.2.

Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs

GPT-5.2 vs GPT-5.6 Terra: questions

Is GPT-5.2 better than GPT-5.6 Terra?
GPT-5.2 (xhigh) and GPT-5.6 Terra (xhigh) are neck and neck on quality (59.9 vs 60.1, within the ±5.7 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 16 Nov 2026, so check the category scores for your use.
GPT-5.2 vs GPT-5.6 Terra: which is cheaper, and will that change?
GPT-5.6 Terra is cheaper today: $4.50 against $4.81 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
Is GPT-5.2 or GPT-5.6 Terra better for coding?
GPT-5.2 scores higher in coding (63 vs 59 on the category index, where 50 is average). On LiveCodeBench, the top shared coding benchmark, it is 85.4% for GPT-5.2 against 85.9% for GPT-5.6 Terra.
Which is faster, GPT-5.2 or GPT-5.6 Terra?
GPT-5.6 Terra streams faster (97 vs 76 output tokens per second, Artificial Analysis 7-day median).
Is GPT-5.2 or GPT-5.6 Terra better for agents and tool use?
GPT-5.6 Terra scores higher in agentic tasks (63 vs 54 on the category index, where 50 is average). On Terminal-Bench Hard, the top shared agents & tools benchmark, it is 47.0% for GPT-5.2 against 62.9% for GPT-5.6 Terra.
More questions (4)
Which is better for long documents?
GPT-5.2 scores higher on long-context benchmarks (66 vs 64). GPT-5.6 Terra accepts more context: 1,050,000 against 400,000 tokens for GPT-5.2.
Which has lower latency?
GPT-5.6 Terra answers first: 19.2 s against 112.5 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
Do independent benchmarks agree on GPT-5.2 vs GPT-5.6 Terra?
Across the 21 benchmarks both have, GPT-5.2 leads 7, GPT-5.6 Terra leads 14. Every overall index that rates both puts GPT-5.6 Terra ahead.
When was GPT-5.6 Terra released?
GPT-5.6 Terra was released on 9 Jul 2026.