Atlas

Compare Models

GPT-6 Astra vs Qwen3.8-Max

GPT-6 Astra leads by 7.6 capability points (≈95% interval +4.9 to +10.2): a clear lead over Qwen3.8-Max · P(higher) ≈ 100%

Wins 31 of 35 shared benchmarks (4 lost)

Capability vs. relative cost
GPT-6 AstraQwen3.8-MaxClosed 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 AstravsQwen3.8-Maxcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins31 – 4 · 35 shared

Comparing matching setups only (263 benchmarks excluded) · suites counted once
← GPT-6 AstraQwen3.8-Max →

GPT-6 AstravsQwen3.8-Maxwins by categoryData Nov 15, 2026Data: Artificial Analysis, Conceptual Reasoning Benchmark Authors, Vercel and 12 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is Qwen3.8-Max minus GPT-6 Astra.

GameBench 2Games94.351.4−42.9
BabaIsBench Stage 0: The IntroGames100.058.0−42.0
Terminal-Bench 4.0 (Artificial Analysis)Code59.618.7−40.9
DeepsecBench — Revalidated V1Code37.815.5−22.3
DeepsecBench — Revalidated V1 (233 issues)Code37.615.5−22.2
AA-Omniscience IndexIndexes43.73.4−20.2
Conceptual Reasoning IndexIndexes73.959.4−20.0
AutomationBench-AA 1.0.6Agents68.549.2−19.3
BuseyBench SVGMultimodal7.35.4−19.0
SCI-AGENT-EXAMScience80.962.7−18.2
LMCAGeneral QA64.446.2−18.1
DeepSWE v1.1Code74.156.6−17.5
Medical Long Context Reasoning (MLCR-AA)Health & Medicine35.019.4−15.6
Artificial Analysis Intelligence Index v4.3.2Indexes52.740.2−12.5
Artificial Analysis Intelligence Index v4.3Indexes52.840.3−12.5
GDP.pdf (Artificial Analysis, All-pass)Professional Work32.220.2−12.0
CritPt (avg@5)Science31.720.0−11.7
Dual Exam Agent IndexIndexes31.320.0−11.3
Humanity's Last ExamGeneral QA54.843.6−11.2
Artificial Analysis Progress Index v1Indexes350320−10.5
ACCoRDGeneral QA81.771.4−10.3
AA-Briefcase EloAgents15621388−10.1
Epoch Capabilities Index (ECI)Indexes169157−9.5
Terminal-Bench 2.1Code89.981.3−8.6
𝜏³-BankingAgents43.151.3+8.2
Ramp Accounting BenchFinance & Accounting48.741.2−7.5
LVBenchMultimodal87.581.8−5.7
DTBenchGeneral QA97.392.0−5.3
Nimble Web ContextSearch72.677.3+4.7
MMMU ProMultimodal86.482.3−4.1
GPQA Diamond (Artificial Analysis avg@5)General QA96.392.7−3.5
SciCode 1.0.1 (Artificial Analysis)Code56.553.2−3.2
GDPval-AA v2Professional Work15801630+2.5
AA-LCR v1.1Search80.078.3−1.7
GDPval-AA v2.1Professional Work15871611+1.1

GPT-6 AstravsQwen3.8-Maxnormalised score marginsData Nov 15, 2026Data: Artificial Analysis, Conceptual Reasoning Benchmark Authors, Vercel and 12 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…