Atlas

Compare Models

GPT-6.2 Astra vs Kimi K3.1 Pro

GPT-6.2 Astra leads by 5.3 capability points (≈95% interval +2.5 to +8.1): a clear lead over Kimi K3.1 Pro · P(higher) ≈ 100%

Wins 29 of 36 shared benchmarks (4 lost, 3 tied)

Capability vs. relative cost
GPT-6.2 AstraKimi K3.1 ProClosed 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-6.2 AstravsKimi K3.1 Procapability vs. relative costData Nov 17, 2026Village LLM Atlasatlas.village.dev

Benchmark wins29 – 4 · 36 shared, 3 tied

Comparing matching setups only (44 benchmarks excluded) · suites counted once
← GPT-6.2 AstraKimi K3.1 Pro →

GPT-6.2 AstravsKimi K3.1 Prowins by categoryData Nov 17, 2026Data: Vals AI, Epoch AI, Artificial Analysis and 3 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Kimi K3.1 Pro minus GPT-6.2 Astra.

Mystery Game PuzzlesGames95.042.0−53.0
Terminal-Bench Science 0.1 (Vals)Science71.431.4−40.0
Terminal-Bench-Science 0.1Science77.138.1−39.0
FrontierMath Tier 4 (v2) - PrivateMath100.065.9−34.1
Furniture Assembly BenchmarkMultimodal89.255.8−33.3
Chess Puzzles (task version 2.0)Games78.045.0−33.0
Humanity's Last Exam DiamondGeneral QA72.239.4−32.8
AA-3D v1.0 EloCode13181112−28.0
Legal Research Bench ImmigrationLegal45.518.2−27.3
MysteryMechanismScience53.226.1−27.0
Vals Science IndexIndexes68.444.4−24.1
Code Migration v1.1 (Vals Index v2.2 subset)Code83.659.7−23.9
SimpleQA Verified (Epoch anti-abstention)General QA78.562.2−16.3
FrontierMath Tiers 1-3 (v2) - PrivateMath92.678.2−14.4
Legal Research Bench CriminalLegal42.128.1−14.0
Legal Research Bench Weighted PassLegal84.171.4−12.6
Legal Research Bench Constitutional/Civil RightsLegal42.131.6−10.5
Epoch Capabilities Index (ECI)Indexes174161−10.2
Legal Research Bench Business/CommercialLegal34.927.9−7.0
Legal Research Bench HealthLegal80.073.3−6.7
Excel Modeling Benchmark v1.1Finance & Accounting76.070.0−6.0
Tax Agent BenchFinance & Accounting57.963.9+6.0
MMMU ProMultimodal89.584.1−5.4
BioMysteryBench (Vals Terminus 2)Science80.775.6−5.2
Vibe Code Bench v1.1Code89.984.9−5.0
HLAB-Vals-1.1Legal16.311.7−4.6
AutomationBench-AA 1.0.6Agents77.273.3−3.9
Agents' Last Exam v1.1 ScoreAgents78.374.4−3.9
Legal Research Bench Administrative/RegulatoryLegal59.156.1−3.0
AA-Briefcase v1.1 EloAgents16441678+1.9
Vals Finance IndexIndexes57.356.0−1.3
Finance Agent v2.1Finance & Accounting60.761.5+0.7
Terminal-Bench 4.0Code63.163.6+0.5
Legal Research Bench Civil LitigationLegal29.629.60.0
Legal Research Bench FamilyLegal13.613.60.0
OTIS Mock AIME 2024-2025Math100.0100.00.0

GPT-6.2 AstravsKimi K3.1 Pronormalised score marginsData Nov 17, 2026Data: Vals AI, Epoch AI, Artificial Analysis and 3 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…