Atlas

Compare Models

GPT-6.2 Astra vs Qwen3.8 Max (0902)

GPT-6.2 Astra leads by 8.7 capability points (≈95% interval +5.9 to +11.5): a clear lead over Qwen3.8 Max (0902) · P(higher) ≈ 100%

Wins 24 of 29 shared benchmarks (5 lost)

Capability vs. relative cost
GPT-6.2 AstraQwen3.8 Max (0902)Closed 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 AstravsQwen3.8 Max (0902)capability vs. relative costData Nov 17, 2026Village LLM Atlasatlas.village.dev

Benchmark wins24 – 5 · 29 shared

Comparing matching setups only (122 benchmarks excluded) · suites counted once
← GPT-6.2 AstraQwen3.8 Max (0902) →

GPT-6.2 AstravsQwen3.8 Max (0902)wins by categoryData Nov 17, 2026Data: Artificial Analysis and Epoch AIVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Qwen3.8 Max (0902) minus GPT-6.2 Astra.

AA-PersonalityBench Slop indexChat & Writing1.073.3+73.1
Terminal-Bench-Science 0.1Science78.611.9−66.7
AA-PersonalityBench Organization indexChat & Writing16.671.4+55.3
Humanity's Last Exam DiamondGeneral QA72.224.0−48.2
AA-PersonalityBench Conviction indexChat & Writing80.935.3−46.1
AA-PersonalityBench Expressiveness indexChat & Writing60.023.8−36.5
CritPt v2 (agentic, avg@5)Science83.751.4−32.3
Artificial Analysis Intelligence Index v5.0.1Indexes62.133.7−28.5
Artificial Analysis Engineering IndexIndexes66.639.3−27.3
FrontierCode 2.0 (Artificial Analysis)Code49.023.4−25.5
Terminal-Bench 4.0 (Artificial Analysis)Code64.138.9−25.3
AutomationBench-AA 1.0.6Agents77.256.2−21.1
AA-Omniscience v1.1 IndexIndexes50.010.5−19.8
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine39.420.0−19.4
AA-PersonalityBench Warmth indexChat & Writing67.848.6−19.4
Artificial Analysis Economics IndexIndexes61.142.5−18.6
Artificial Analysis Legal IndexIndexes58.741.0−17.7
EnterpriseOps-Gym-AAAgents29.947.6+17.7
AA-AnalystAgent (pass^5)Professional Work62.545.0−17.5
AA-PersonalityBench Informativeness indexChat & Writing22.039.1+17.3
Artificial Analysis Healthcare & Medical IndexIndexes57.040.8−16.2
Artificial Analysis Strategy & Ops IndexIndexes63.347.3−16.0
Artificial Analysis Finance & Accounting IndexIndexes60.945.0−15.9
Epoch Capabilities Index (ECI)Indexes174156−14.5
AA-PersonalityBench Imagination indexChat & Writing60.448.6−11.9
AA-PersonalityBench Transparency indexChat & Writing36.144.8+8.8
GDP.pdf (Artificial Analysis, All-pass)Professional Work29.822.8−7.0
MMMU ProMultimodal89.582.8−6.7
AA-Briefcase v1.1 EloAgents16441604−2.2

GPT-6.2 AstravsQwen3.8 Max (0902)normalised score marginsData Nov 17, 2026Data: Artificial Analysis and Epoch AIVillage 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…