Atlas

Compare Models

GPT-6.1 Terra vs GPT-6 Sol

GPT-6.1 Terra leads by 1.5 capability points (≈95% interval −0.9 to +3.9): statistically tied · P(higher) ≈ 89%

Wins 61 of 99 shared benchmarks (36 lost, 2 tied)

Capability vs. relative cost
GPT-6.1 TerraGPT-6 SolClosed 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 Solcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins61 – 36 · 99 shared, 2 tied

Comparing matching setups only (1 benchmark excluded) · suites counted once
← GPT-6.1 TerraGPT-6 Sol →

GPT-6.1 TerravsGPT-6 Solwins by categoryData Nov 15, 2026Data: Artificial Analysis, benchmarks.bio, Epoch AI and 29 othersVillage LLM Atlasatlas.village.dev

Margins

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

Vending-Bench 2Games910714428+32.2
Terminal-Bench-Science 0.1Science57.630.0−27.6
EBR-bench v4 (card ban, single agent)Games78.153.3−24.8
Furniture Assembly BenchmarkMultimodal79.258.3−20.8
ExploitBench - Internal Port (June-August 2026)Cybersecurity24.55.5−19.0
Mystery Game PuzzlesGames74.056.0−18.0
SpireClimbGames22.39.2−13.2
TxBench Oligo Discovery V1 (113 evaluations)Science30.443.1+12.7
HardGeoBench Avg DistanceMultimodal10911497−12.5
HardGeoBench Country AccuracyMultimodal58.846.3−12.5
BioSecBench-Structure V1 (58 evaluations)Science25.913.8−12.1
Terminal-Bench 4.0 (Artificial Analysis)Code55.643.9−11.6
Bug Hunt BenchCode38.727.9−10.8
APEX-Agents 1.1Professional Work79.368.7−10.6
ScBench-Long V1 Clean RerunsScience18.828.4+9.7
ChartographyMultimodal63.253.6−9.6
LMArena Agent Arena Net Improvement (Steering Burden)Agents12.06.9−9.6
SpatialBench-Long V1 Clean RerunsScience40.531.6−8.9
HardGeoBench <=100miMultimodal46.337.5−8.7
HardGeoBench <=10miMultimodal35.026.3−8.7
NoCoT-Bench NCRI 15.2General QA130115−8.7
ALE-BenchCode28752637−8.4
WeirdML v3Science28.119.7−8.4
MacroscopeBenchCode82.374.3−8.0
BioSecBench-Surveillance V1 Clean RerunsScience39.947.7+7.8
GBENCH One-shot Coding GScoreGames77.670.1−7.5
TxBench Preclinical Pharmacology V1 Clean RerunsScience55.763.0+7.3
Artificial Analysis Intelligence Index v5.0.1Indexes51.344.2−7.2
MazeBench ASCIIGames8.01.0−7.0
FrontierCode 2.0 (Artificial Analysis)Code42.335.6−6.7
GameBench 2Games85.879.4−6.4
LLM Chess EloGames13911545+6.3
NoCoT-Bench NCKI (kspine v3)General QA110119+6.3
τ³-Banking (Mercor run, bm25_grep retrieval)Agents37.143.0+5.9
SpatialBench V1 Clean RerunsScience73.367.5−5.8
VoxelBench Text RatingGames23132394+5.8
Blueprint-Bench 2Multimodal0.40.4−5.2
Terminal-Bench 4.0Code49.544.4−5.1
LMArena Agent Arena Praise vs ComplaintAgents20.010.0−5.0
FORTRESS Average Risk Score (ARS)Safety18.213.5+4.7
HealthBench ProfessionalHealth & Medicine64.259.5−4.7
AA-Briefcase v1.1 EloAgents14011476+4.1
APEX-Accounting 1.1Finance & Accounting63.759.7−4.0
TaxCalcBench TY25Finance & Accounting24.028.0+4.0
HealthBench (updated implementation)Health & Medicine51.147.1−4.0
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine16.020.0+4.0
VariantBench V1 Clean RerunsScience34.838.7+4.0
Artificial Analysis Progress Index v1Indexes345334−3.9
HealthBench Hard (updated implementation)Health & Medicine26.022.1−3.9
Vibe Code Bench v1.1Code84.087.8+3.8
BioSecBench-Refusal V1 Clean RerunsSafety34.237.9+3.7
ChessBench CoherenceGames84.888.2+3.4
GDP.pdf (Artificial Analysis, All-pass)Professional Work28.625.2−3.4
SimpleQA Verified (Epoch anti-abstention)General QA57.360.7+3.4
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine30.633.9+3.3
MetagenomicsBench V1Science38.742.0+3.3
Epoch Capabilities Index (ECI)Indexes168164−3.3
ScBench V1 Clean RerunsScience58.361.5+3.2
LMArena Agent Arena Steering BurdenAgents16.610.1−3.2
VulcanBench Frontier v4Code90.086.8−3.2
ChessBench EloGames12511303+3.1
Toloka Arena (v3)Agents55.058.0+3.0
Legal Research BenchLegal26.028.8+2.9
CritPt (avg@5)Science28.030.9+2.9
FrontierMath Tier 4 (v2) - PrivateMath92.790.0−2.7
TxBench Antibody Discovery V1 (100 evaluations)Science40.743.3+2.7
LMArena Agent Arena Confirmed SuccessAgents9.74.5−2.6
SciCode 1.0.1 (Artificial Analysis)Code55.157.6+2.5
BuseyBench SVGMultimodal7.37.5+2.5
LiveBench 2.0Agents79.781.9+2.3
MMMU ProMultimodal84.782.5−2.1
Humanity's Last Exam DiamondGeneral QA47.345.2−2.1
DistressBenchSafety65.968.0+2.1
ChessBenchGames64.062.0−2.0
DualEntry Accounting AI BenchmarkFinance & Accounting78.276.2−2.0
Tax Agent BenchFinance & Accounting55.053.0−2.0
HealthBench (length-adjusted)Health & Medicine55.153.2−1.9
GDPval-AA v2.1Professional Work14761514+1.7
AutomationBench-AA 1.0.6Agents60.261.7+1.5
AA-Omniscience IndexIndexes29.827.1−1.3
MentalHealthBenchSafety55.554.2−1.3
HealthBench Hard (length-adjusted)Health & Medicine31.430.1−1.3
EpiBench V1 Clean RerunsScience23.322.0−1.3
LMArena Agent Arena Bash ReliabilityAgents13.410.9−1.2
Design Arena Overall Code CategoriesCode13001293−1.2
FrontierMath Tiers 1-3 (v2) - PrivateMath90.989.8−1.1
AA-LCR v1.1Search82.783.7+1.0
HealthBench Consensus (updated implementation)Health & Medicine94.895.8+1.0
ChessBench AccuracyGames78.377.4−1.0
LMArena Code Arena WebDevCode16791673−0.7
APEX-1 (Mercor, mean score)Professional Work64.263.6−0.6
CritPt v2 (agentic, avg@5)Science81.480.9−0.6
Artificial Analysis Intelligence Index v4.3.2Indexes48.147.6−0.5
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal4.54.1−0.4
BioSecBench-Function V1 Clean RerunsScience39.039.3+0.3
Humanity's Last Exam (Text-Only)General QA46.246.30.0
LMArena Agent Arena Tool HallucinationAgents1.01.00.0
FrontierPhysics v0.1Science14.314.30.0
OTIS Mock AIME 2024-2025Math100.0100.00.0

GPT-6.1 TerravsGPT-6 Solnormalised score marginsData Nov 15, 2026Data: Artificial Analysis, benchmarks.bio, Epoch AI and 29 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…