The Model Index
Up to four models

GPT-5.6 Terra vs Kimi K3

On the index (GPT-4 = 100), GPT-5.6 Terra scores 172 and Kimi K3 168. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.

Next to GPT-5.6 Terra
Presets
GPT-5.6 Terra
172156–188
WeightsClosed
ReleasedJul 2026
Price in / out$2.0 / $12
Context1.05M
Core tests taken6 of 9
Kimi K3
168152–183
WeightsOpen
ReleasedJul 2026
Price in / out$0.19 / $16
Context1.05M
Core tests taken6 of 9

GPT-5.6 Terra and Kimi K3 are too close to call (4 points apart, within the uncertainty), ahead on 6 of 6 shared core tests, at a similar price.

Test by testthird-party results

GPT-5.6 TerraKimi K3
0%25%50%75%100%REASONING & KNOWLEDGEGPQA Diamond (Epoch AI)GPQA Diamond (Epoch AI)LiveBench Data AnalysisLiveBench Data AnalysisLiveBench IFLiveBench IFLiveBench LanguageLiveBench LanguageLiveBench ReasoningLiveBench ReasoningGPQA Diamond (Vals AI)GPQA Diamond (Vals AI)MMLU ProMMLU ProMMMU ProMMMU ProSimpleBenchSimpleBenchSimpleQA VerifiedSimpleQA VerifiedMATHSFrontierMath (tiers 1-3)FrontierMath (tiers 1-3)AIME-style maths (OTIS mock)AIME-style maths (OTIS mock)LiveBench MathematicsLiveBench MathematicsProofBench v1.1ProofBench v1.1Chess puzzlesChess puzzlesFrontierMath (tier 4)FrontierMath (tier 4)Surface Evolver BenchSurface Evolver BenchCODINGTerminal-Bench 4.0Terminal-Bench 4.0LiveBench Agentic CodingLiveBench Agentic CodingLiveBench CodingLiveBench CodingCode MigrationCode MigrationCyberBench v1.1CyberBench v1.1IOIIOILiveCodeBenchLiveCodeBenchProgramBenchProgramBenchSWE-benchSWE-benchVibe Code Bench 1-100Vibe Code Bench 1-100Vibe Code Bench v1.1Vibe Code Bench v1.1DeepSWEDeepSWEMirrorCodeMirrorCodeWeirdMLWeirdMLAGENTSAPEX-AgentsAPEX-AgentsTerminal-Bench-ScienceTerminal-Bench-ScienceDTBenchDTBenchFurniture assemblyFurniture assemblyLMCALMCAMystery game puzzlesMystery game puzzlesPROFESSIONAL WORKFinance Agent (v2)Finance Agent (v2)CorpFin v2CorpFin v2EMBEMBLegalBenchLegalBenchLegal Research BenchLegal Research BenchMedCodeMedCodeMedScribeMedScribeMortgageTaxMortgageTaxPublic Benefits Bench v1.1Public Benefits Bench v1.1SAGESAGETax Agent BenchTax Agent BenchTaxEval v2TaxEval v2NOVEL PROBLEMSARC-AGI-2ARC-AGI-2ARC-AGI-1ARC-AGI-1
Every result as a table
TestGPT-5.6 TerraKimi K3
Reasoning & knowledge
GPQA Diamond (Epoch AI)93.3%93.1%
LiveBench Data Analysis79.3%78.7%
LiveBench IF64.6%71.4%
LiveBench Language82.9%85.5%
LiveBench Reasoning90.6%90.7%
GPQA Diamond (Vals AI)90.9%92.9%
MMLU Pro86.7%88.0%
MMMU Pro86.5%88.1%
SimpleBench48.9%60.7%
SimpleQA Verified43.2%50.6%
Maths
FrontierMath (tiers 1-3)86.0%72.2%
AIME-style maths (OTIS mock)99.7%97.2%
LiveBench Mathematics94.9%84.4%
ProofBench v1.174.0%87.0%
Chess puzzles47.0%37.0%
FrontierMath (tier 4)70.7%39.0%
Surface Evolver Bench83.8%95.0%
Coding
Terminal-Bench 4.022.7%17.2%
LiveBench Agentic Coding54.9%62.2%
LiveBench Coding78.3%81.5%
Code Migration47.8%16.1%
CyberBench v1.172.1%75.2%
IOI87.6%48.9%
LiveCodeBench85.9%87.2%
ProgramBench0.5%2.0%
SWE-bench95.4%93.4%
Vibe Code Bench 1-10014.8%18.2%
Vibe Code Bench v1.174.6%85.0%
DeepSWE69.6%68.5%
MirrorCode41.1%54.4%
WeirdML78.3%82.6%
Agents
APEX-Agents58.2%47.8%
Terminal-Bench-Science10.0%1.4%
DTBench93.3%91.2%
Furniture assembly54.2%34.2%
LMCA54.9%52.7%
Mystery game puzzles35.0%26.0%
Professional work
Finance Agent (v2)54.4%53.1%
CorpFin v265.3%71.6%
EMB66.2%66.7%
LegalBench85.1%86.2%
Legal Research Bench41.3%46.2%
MedCode43.4%49.4%
MedScribe82.9%88.0%
MortgageTax67.3%66.3%
Public Benefits Bench v1.162.4%68.2%
SAGE87.0%86.8%
Tax Agent Bench30.0%33.4%
TaxEval v276.2%75.7%
Novel problems
ARC-AGI-283.9%60.4%
ARC-AGI-196.5%94.5%

Tests at least two of these models have taken. Bold rows are the core tests.

This page as a chartDownload PNG