Atlas

Compare Models

GPT-6.1 Sol vs GPT-6.1 Terra

GPT-6.1 Sol leads by 1.0 capability points (≈95% interval −1.6 to +3.5): statistically tied · P(higher) ≈ 77%

Wins 104 of 134 shared benchmarks (25 lost, 5 tied)

Capability vs. relative cost
GPT-6.1 SolGPT-6.1 TerraClosed 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 SolvsGPT-6.1 Terracapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins104 – 25 · 134 shared, 5 tied

Comparing matching setups only (2 benchmarks excluded) · suites counted once
← GPT-6.1 SolGPT-6.1 Terra →

GPT-6.1 SolvsGPT-6.1 Terrawins by categoryData Nov 15, 2026Data: Artificial Analysis, benchmarks.bio, Vals AI and 33 othersVillage LLM Atlasatlas.village.dev

Margins

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

EBR-bench v4 (card ban, single agent)Games54.378.1+23.8
TxBench Oligo Discovery V1 (113 evaluations)Science52.530.4−22.1
ChessBench EloGames15621251−18.5
VoxelBench Text RatingGames25502313−16.9
SimpleQA Verified (Epoch anti-abstention)General QA73.957.3−16.6
TaxCalcBench TY25Finance & Accounting38.024.0−14.0
NoCoT-Bench NCKI (kspine v3)General QA130110−13.8
Legal Research BenchLegal38.526.0−12.5
Humanity's Last Exam DiamondGeneral QA58.847.3−11.5
VariantBench V1 Clean RerunsScience46.034.8−11.3
Visa Agentic Commerce BenchmarkAgents62.251.1−11.1
Toloka Arena (v3)Agents66.055.0−11.0
BioSecBench-Surveillance V1 Clean RerunsScience50.839.9−10.9
NoCoT-Bench NCRI 15.2General QA149130−10.2
WeirdML v3Science38.328.1−10.2
Vals Legal IndexIndexes68.758.8−10.0
Design Arena Overall Code CategoriesCode13561300−9.3
LLM Chess EloGames16121391−9.0
TxBench Antibody Discovery V1 (100 evaluations)Science49.740.7−9.0
LMArena Code Arena WebDevCode17431679−8.4
AA-Briefcase v1.1 EloAgents15551401−8.4
DistressBenchSafety74.265.9−8.3
SWE Atlas v1.1 - OverallCode65.957.7−8.2
OncologyBench Strict (25 tasks, October 9 rubric)Health & Medicine24.016.0−8.0
ScBench-Long V1 Clean RerunsScience26.718.8−8.0
Artificial Analysis Economics IndexIndexes56.348.5−7.8
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine38.330.6−7.8
MultiChallengeChat & Writing82.074.2−7.7
Artificial Analysis Strategy & Ops IndexIndexes57.650.0−7.6
HealthBench Hard (updated implementation)Health & Medicine33.426.0−7.4
FrontierMath Tier 4 (v2) - PrivateMath100.092.7−7.3
Artificial Analysis Healthcare & Medical IndexIndexes52.044.7−7.3
Artificial Analysis Legal IndexIndexes53.746.4−7.3
Tax Agent BenchFinance & Accounting62.355.0−7.3
Artificial Analysis Finance & Accounting IndexIndexes56.249.0−7.2
DeepFinanceFinance & Accounting54.347.2−7.1
BioSecBench-Function V1 Clean RerunsScience46.039.0−7.0
MetagenomicsBench V1Science45.738.7−7.0
BuseyBench SVGMultimodal8.07.3−6.7
AutomationBench-AA 1.0.6Agents66.660.2−6.4
Humanity's Last Exam (Text-Only)General QA52.646.2−6.4
HardGeoBench Country AccuracyMultimodal52.558.8+6.3
GameBench 2Games92.085.8−6.1
APEX-Agents 1.1Professional Work73.279.3+6.1
SpatialBench V1 Clean RerunsScience67.373.3+6.1
Artificial Analysis Coding Agent Index v2.0Indexes54.148.1−6.0
Mystery Game PuzzlesGames80.074.0−6.0
τ³-Banking (Mercor run, bm25_grep retrieval)Agents43.037.1−5.9
AA-Omniscience IndexIndexes41.529.8−5.9
Finance Agent v2.1Finance & Accounting61.856.0−5.9
GDPval-AA v2.1Professional Work16071476−5.8
LiveBench 2.0Agents85.479.7−5.7
HiL-Bench (Human-in-Loop Benchmark)Agents62.368.0+5.7
HealthBench (updated implementation)Health & Medicine56.751.1−5.6
MacroscopeBenchCode76.782.3+5.6
Terminal-Bench 4.0Code55.149.5−5.6
ScBench V1 Clean RerunsScience63.658.3−5.3
FrontierPhysics v0.1Science19.614.3−5.3
EpiBench V1 Clean RerunsScience28.323.3−5.0
TxBench-Antibody-Development V1 (60 evaluations)Science27.822.8−5.0
FrontierCode 1.1 Main (Artificial Analysis)Code46.341.4−4.9
Vibe Code Bench v1.1Code88.984.0−4.9
HealthBench Hard (length-adjusted)Health & Medicine36.231.4−4.8
ChessBenchGames68.764.0−4.7
Artificial Analysis Engineering IndexIndexes59.454.7−4.6
BioSecBench-Structure V1 (58 evaluations)Science21.325.9+4.6
HLAB-Vals-1.1Legal6.31.7−4.6
HardGeoBench Avg DistanceMultimodal12391091+4.6
FrontierMemory v1.1Agents51.146.7−4.4
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal8.84.5−4.3
LMArena Agent Arena Bash ReliabilityAgents5.013.4+4.2
ALE-BenchCode29902875−4.1
TxBench Preclinical Pharmacology V1 Clean RerunsScience59.755.7−4.0
MKA (Artificial Analysis)Professional Work47.743.8−4.0
LMArena Agent Arena Praise vs ComplaintAgents27.520.0−3.7
Artificial Analysis Intelligence Index v4.3.2Indexes51.848.1−3.7
CritPt (avg@5)Science31.728.0−3.7
Artificial Analysis Intelligence Index v5.0Indexes54.551.0−3.5
Artificial Analysis Coding Agent Index v2.1Indexes52.248.6−3.5
Bug Hunt BenchCode42.238.7−3.5
Vals Index v2.2Indexes63.960.5−3.4
TissueMicroscopyBench V1 (49 evaluations)Science22.425.9+3.4
BioSecBench-Refusal V1 Clean RerunsSafety37.634.2−3.4
GDP.pdf (Artificial Analysis, All-pass)Professional Work32.028.6−3.4
HealthBench (length-adjusted)Health & Medicine58.555.1−3.4
Artificial Analysis Intelligence Index v5.0.1Indexes54.651.3−3.3
DualEntry Accounting AI BenchmarkFinance & Accounting81.278.2−3.0
ExploitBench - Internal Port (June-August 2026)Cybersecurity21.524.5+3.0
HealthBench ProfessionalHealth & Medicine67.264.2−3.0
Code Migration v1.1 (Vals Index v2.2 subset)Code75.172.2−3.0
FrontierMath Tiers 1-3 (v2) - PrivateMath93.790.9−2.8
ChessBench CoherenceGames87.584.8−2.7
ChessBench AccuracyGames80.778.3−2.6
HardGeoBench <=100miMultimodal43.846.3+2.5
Vals Finance IndexIndexes55.853.3−2.5
MentalHealthBenchSafety57.955.5−2.4
Excel Modeling Benchmark v1.1Finance & Accounting71.173.3+2.2
RareDiseaseBench V1 (35 evaluations)Science27.625.7−1.9
FrontierCode 2.0 (Artificial Analysis)Code40.642.3+1.7
VulcanBench Frontier v4Code88.390.0+1.6
LMArena Agent Arena Steering BurdenAgents13.416.6+1.6
SpireClimbGames23.822.3−1.5
GBENCH One-shot Coding GScoreGames79.077.6−1.4
Agents' Last Exam Overall ScoreAgents58.660.0+1.4
AA-LCR v1.1Search84.082.7−1.3
VIALS-AA (pass^3)Multimodal59.060.2+1.2
VIALS-AA (average of 3)Multimodal64.866.0+1.2
BioSecBench-RespiratoryVirus-Surveillance V1 (110 evaluations)Science12.411.2−1.2
CritPt v2 (agentic, avg@5)Science82.681.4−1.1
Artificial Analysis Progress Index v1Indexes348345−1.1
HealthBench Consensus (updated implementation)Health & Medicine95.994.8−1.1
DAXBench v2Code92.693.6+1.1
LMArena Agent Arena Confirmed SuccessAgents11.79.7−1.0
MazeBench ASCIIGames9.08.0−1.0
LMArena Agent Arena Net Improvement (Steering Burden)Agents12.512.0−0.9
Furniture Assembly BenchmarkMultimodal80.079.2−0.8
SciCode 1.0.1 (Artificial Analysis)Code55.855.1−0.7
VIALS-AA (pass@3)Multimodal70.871.4+0.6
Epoch Capabilities Index (ECI)Indexes169168−0.6
Android Bench 2.0Code29.328.7−0.6
APEX-Accounting 1.1Finance & Accounting64.363.7−0.6
SpatialBench-Long V1 Clean RerunsScience39.940.5+0.6
Terminal-Bench 4.0 (Artificial Analysis)Code56.155.6−0.5
Terminal-Bench-Science 0.1Science58.157.6−0.5
MMMU ProMultimodal85.184.7−0.4
Prophet Arena Arena Score (Fixed Context)Forecasting84.885.2+0.4
NeuroscienceBench V1 (91 evaluations)Science22.022.3+0.4
Prophet Arena Arena Score (Agentic, OpenCode)Forecasting86.085.7−0.3
GDPxProfessional Work8.48.50.0
LMArena Agent Arena Tool HallucinationAgents1.01.00.0
Cua Driver Bench v2 (Linux, background)Agents12.512.50.0
Cua Driver Bench v2 (Linux, foreground)Agents52.052.00.0
HardGeoBench <=10miMultimodal35.035.00.0
OTIS Mock AIME 2024-2025Math100.0100.00.0

GPT-6.1 SolvsGPT-6.1 Terranormalised score marginsData Nov 15, 2026Data: Artificial Analysis, benchmarks.bio, Vals AI and 33 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…