Atlas

Compare Models

GPT-6.1 Terra vs GPT-6 Astra

GPT-6 Astra leads by 1.9 capability points (≈95% interval −0.6 to +4.5): statistically tied · P(higher) ≈ 93%

Wins 115 of 133 shared benchmarks (16 lost, 2 tied)

Capability vs. relative cost
GPT-6.1 TerraGPT-6 AstraClosed weightsOpen weights

Each line runs through one model’s priced reasoning settings, cheapest to most expensive; grey points are other models’ priced settings in the same range. Results whose reasoning effort the source didn’t report are left out wherever a model has named effort levels. Capability and relative cost are the published estimates (relative cost is cost per task as a multiple of GPT-5 (high)); hover a setting, or use the arrow keys, to set it beside the other model’s same setting.

GPT-6.1 TerravsGPT-6 Astracapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins16 – 115 · 133 shared, 2 tied

Comparing matching setups only (2 benchmarks excluded) · suites counted once
← GPT-6.1 TerraGPT-6 Astra →

GPT-6.1 TerravsGPT-6 Astrawins by categoryData Nov 15, 2026Data: Artificial Analysis, benchmarks.bio, Epoch AI and 34 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-6 Astra minus GPT-6.1 Terra.

ChessBench EloGames12511755+30.0
VoxelBench Text RatingGames23132625+22.2
TxBench Oligo Discovery V1 (113 evaluations)Science30.451.9+21.5
TaxCalcBench TY25Finance & Accounting24.044.0+20.0
SimpleQA Verified (Epoch anti-abstention)General QA57.375.6+18.3
Humanity's Last Exam DiamondGeneral QA47.363.8+16.5
NoCoT-Bench NCRI 15.2General QA130159+15.7
TxBench Antibody Discovery V1 (100 evaluations)Science40.756.0+15.3
NoCoT-Bench NCKI (kspine v3)General QA110131+14.8
WeirdML v3Science28.142.2+14.1
Legal Research BenchLegal26.039.4+13.5
Visa Agentic Commerce BenchmarkAgents51.164.4+13.3
Terminal-Bench 4.0 (Artificial Analysis)Code46.559.6+13.1
Terminal-Bench-Science 0.1Science41.954.8+12.9
ChessBench CoherenceGames84.897.0+12.2
Toloka Arena (v3)Agents55.067.0+12.0
VariantBench V1 Clean RerunsScience34.846.6+11.9
SWE Atlas v1.1 - OverallCode57.769.4+11.7
SpireClimbGames22.334.0+11.7
ALE-BenchCode28433164+11.4
LMArena Code Arena WebDevCode16791764+11.2
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine30.641.7+11.1
MetagenomicsBench V1Science38.749.7+11.0
FrontierCode 1.1 Main (Artificial Analysis)Code36.047.0+11.0
DeepFinanceFinance & Accounting47.257.6+10.4
Terminal-Bench 4.0Code49.559.6+10.1
Mystery Game PuzzlesGames74.084.0+10.0
TxBench-Antibody-Development V1 (60 evaluations)Science22.832.8+10.0
Vals Legal IndexIndexes58.868.4+9.6
BioSecBench-Function V1 Clean RerunsScience39.048.6+9.6
τ³-Banking (Mercor run, bm25_grep retrieval)Agents37.146.7+9.6
Artificial Analysis Legal IndexIndexes46.455.6+9.2
AA-Briefcase v1.1 EloAgents14011569+9.2
LLM Chess EloGames13911614+9.1
Artificial Analysis Healthcare & Medical IndexIndexes44.753.7+9.0
FrontierPhysics v0.1Science14.323.2+8.9
Artificial Analysis Economics IndexIndexes48.557.2+8.7
BioSecBench-Surveillance V1 Clean RerunsScience39.948.5+8.6
GameBench 2Games85.894.3+8.5
ScBench-Long V1 Clean RerunsScience18.827.3+8.5
DistressBenchSafety65.974.3+8.4
Artificial Analysis Finance & Accounting IndexIndexes49.057.4+8.4
AutomationBench-AA 1.0.6Agents60.268.5+8.3
Tax Agent BenchFinance & Accounting55.063.3+8.3
HealthBench Hard (updated implementation)Health & Medicine26.034.2+8.2
MazeBench ASCIIGames8.016.0+8.0
Artificial Analysis Strategy & Ops IndexIndexes50.058.0+8.0
ChartographyMultimodal63.271.0+7.8
RareDiseaseBench V1 (35 evaluations)Science25.733.3+7.6
Blueprint-Bench 2Multimodal0.40.5+7.6
ChessBench AccuracyGames78.385.0+7.2
Artificial Analysis Engineering IndexIndexes54.761.8+7.1
ExploitBench - Internal Port (June-August 2026)Cybersecurity24.531.5+7.0
AA-Omniscience IndexIndexes29.843.7+7.0
Humanity's Last Exam (Text-Only)General QA46.253.1+6.8
BioSecBench-Refusal V1 Clean RerunsSafety34.240.8+6.6
Artificial Analysis Coding Agent Index v2.0Indexes48.154.7+6.6
MKA (Artificial Analysis)Professional Work43.850.2+6.5
LiveBench 2.0Agents79.786.1+6.5
TissueMicroscopyBench V1 (49 evaluations)Science25.919.7−6.1
FORTRESS Average Risk Score (ARS)Safety18.212.1+6.0
ChessBenchGames64.070.0+6.0
Code Migration v1.1 (Vals Index v2.2 subset)Code72.278.0+5.8
LMArena Agent Arena Praise vs ComplaintAgents20.031.7+5.8
HealthBench (updated implementation)Health & Medicine51.156.9+5.8
SpatialBench V1 Clean RerunsScience73.367.5−5.8
HardGeoBench Avg DistanceMultimodal1091908+5.7
ScBench V1 Clean RerunsScience58.363.9+5.6
Artificial Analysis Intelligence Index v5.0.1Indexes51.356.9+5.5
HLAB-Vals-1.1Legal1.77.1+5.4
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine16.021.3+5.3
HealthBench Hard (length-adjusted)Health & Medicine31.436.6+5.2
APEX-1 (Mercor, mean score)Professional Work64.269.3+5.1
EpiBench V1 Clean RerunsScience23.328.3+5.0
HardGeoBench <=100miMultimodal46.351.3+5.0
FrontierMath Tier 4 (v2) - PrivateMath92.797.6+4.9
GDPval-AA v2.1Professional Work14761587+4.9
LMArena Agent Arena Net Improvement (Steering Burden)Agents12.014.5+4.8
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal4.59.2+4.7
BioSecBench-Structure V1 (58 evaluations)Science25.921.3−4.6
Artificial Analysis Intelligence Index v4.3.2Indexes48.152.7+4.6
Finance Agent v2.1Finance & Accounting56.060.4+4.5
Artificial Analysis Coding Agent Index v2.1Indexes48.653.1+4.5
MacroscopeBenchCode82.378.0−4.3
Artificial Analysis Intelligence Index v5.0Indexes51.055.3+4.3
APEX-Agents 1.1Professional Work79.375.1−4.2
Bug Hunt BenchCode38.742.9+4.1
HealthBench ProfessionalHealth & Medicine64.268.2+4.0
CritPt (avg@5)Science28.031.7+3.7
TxBench Preclinical Pharmacology V1 Clean RerunsScience55.759.3+3.7
GDP.pdf (Artificial Analysis, All-pass)Professional Work28.632.2+3.6
MentalHealthBenchSafety55.558.7+3.2
HealthBench (length-adjusted)Health & Medicine55.158.3+3.2
DAXBench v2Code93.696.8+3.2
DualEntry Accounting AI BenchmarkFinance & Accounting78.281.2+3.0
FrontierMath Tiers 1-3 (v2) - PrivateMath90.993.7+2.8
LMArena Agent Arena Bash ReliabilityAgents13.47.8−2.8
HiL-Bench (Human-in-Loop Benchmark)Agents68.065.3−2.7
AA-LCR v1.1Search82.780.0−2.7
HardGeoBench <=10miMultimodal35.037.5+2.5
VIALS-AA (pass^3)Multimodal60.262.7+2.5
Vals Finance IndexIndexes53.355.6+2.3
VIALS-AA (average of 3)Multimodal66.068.3+2.3
LMArena Agent Arena Confirmed SuccessAgents9.714.0+2.2
BioSecBench-RespiratoryVirus-Surveillance V1 (110 evaluations)Science11.213.3+2.1
EBR-bench v4 (card ban, single agent)Games78.176.2−1.9
VIALS-AA (pass@3)Multimodal71.473.3+1.9
MMMU ProMultimodal84.786.4+1.7
CritPt v2 (agentic, avg@5)Science81.483.1+1.7
GBENCH One-shot Coding GScoreGames77.679.2+1.7
Artificial Analysis Progress Index v1Indexes345350+1.5
GPQA DiamondGeneral QA94.395.8+1.5
SciCode 1.0.1 (Artificial Analysis)Code55.156.5+1.4
HardGeoBench Country AccuracyMultimodal58.860.0+1.3
SpatialBench-Long V1 Clean RerunsScience40.539.3−1.2
GDPxProfessional Work8.59.6+1.1
APEX-Accounting 1.1Finance & Accounting63.764.7+1.0
LMArena Agent Arena Steering BurdenAgents16.614.8−0.9
Epoch Capabilities Index (ECI)Indexes168169+0.8
Furniture Assembly BenchmarkMultimodal79.280.0+0.8
Agents' Last Exam v1.1 (ALE v1.1)Agents41.842.6+0.8
Agents' Last Exam Overall ScoreAgents60.059.3−0.7
Android Bench 2.0Code28.728.0−0.7
VulcanBench Frontier v4Code90.089.3−0.7
Excel Modeling Benchmark v1.1Finance & Accounting73.372.7−0.6
HealthBench Consensus (updated implementation)Health & Medicine94.895.4+0.6
Prophet Arena Arena Score (Fixed Context)Forecasting85.284.7−0.5
Prophet Arena Arena Score (Agentic, OpenCode)Forecasting85.786.2+0.5
BuseyBench SVGMultimodal7.37.3+0.4
NeuroscienceBench V1 (91 evaluations)Science22.322.7+0.4
FrontierCode 2.0 (Artificial Analysis)Code42.342.4+0.1
LMArena Agent Arena Tool HallucinationAgents1.01.00.0
OTIS Mock AIME 2024-2025Math100.0100.00.0

GPT-6.1 TerravsGPT-6 Astranormalised score marginsData Nov 15, 2026Data: Artificial Analysis, benchmarks.bio, Epoch AI and 34 othersVillage LLM Atlasatlas.village.dev

Win counts use each benchmark’s declared score direction. Represented suite slices and the derived Capability score are excluded from headline totals. Missing or incompatible setups are excluded, never counted as losses.
Loading Atlas data…