Atlas

Compare Models

Gemini 3.8 Flash vs Qwen3.8 Max (0902)

Gemini 3.8 Flash leads by 0.4 capability points (≈95% interval −2.0 to +2.7): statistically tied · P(higher) ≈ 62%

Benchmark wins are even: 25 each of 51 shared.

Capability vs. relative cost
Gemini 3.8 FlashQwen3.8 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.

Gemini 3.8 FlashvsQwen3.8 Max (0902)capability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins25 – 25 · 51 shared, 1 tied

Comparing matching setups only (157 benchmarks excluded) · suites counted once
← Gemini 3.8 FlashQwen3.8 Max (0902) →

Gemini 3.8 FlashvsQwen3.8 Max (0902)wins by categoryData Nov 15, 2026Data: Artificial Analysis, AIMING Lab (UNC-Chapel Hill), BuseyBench and 8 othersVillage 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 Gemini 3.8 Flash.

AA-PersonalityBench Expressiveness indexChat & Writing79.123.8−55.8
AA-PersonalityBench Conviction indexChat & Writing84.835.3−50.0
AA-PersonalityBench Informativeness indexChat & Writing81.039.1−42.3
Last Translation Benchmark (LTBv1-eval, blind)Chat & Writing39.812.5−27.3
AA-Briefcase EloAgents12021622+24.4
AA-PersonalityBench Transparency indexChat & Writing23.844.8+21.2
Terminal-Bench 4.0 (Artificial Analysis)Code19.738.9+19.2
Terminal-Bench 4.0Code13.127.0+13.8
CritPt v2 (agentic, avg@5)Science65.151.4−13.7
ITBench-AAAgents52.540.3−12.3
AA-PersonalityBench Warmth indexChat & Writing60.048.6−11.5
GDPval-AA v2Professional Work14641689+11.1
GDPval-AA v2.1Professional Work14421674+10.2
Code2Video-AA EloCode8991021+9.9
VIALS-AA (pass^3)Multimodal41.031.7−9.3
AA-Omniscience IndexIndexes29.612.0−8.8
VIALS-AA (pass@3)Multimodal64.072.7+8.7
BuseyBench SVGMultimodal6.65.8−8.4
AA-PersonalityBench Imagination indexChat & Writing41.048.6+7.7
Visa Agentic Commerce BenchmarkAgents53.360.0+6.7
FrontierCode 1.1 Main (Artificial Analysis)Code37.143.2+6.0
Veracode GenAI Code Security (Fall 2026)Cybersecurity62.068.0+6.0
RSI-Exam 0.1 (Full 88)Agents0.30.4+5.2
Humanity's Last Exam DiamondGeneral QA29.124.0−5.1
AutomationBench-AA 1.0.6Agents60.956.2−4.7
FrontierCode 2.0 (Artificial Analysis)Code18.723.4+4.7
SciCode 1.0.1 (Artificial Analysis)Code56.652.1−4.5
Artificial Analysis Intelligence Index v4.3.2Indexes40.945.4+4.5
DAXBench v2Code87.283.0−4.3
Artificial Analysis Intelligence Index v4.3Indexes41.245.4+4.2
AA-LCR v1.1Search84.080.3−3.7
Artificial Analysis Intelligence Index v5.0Indexes30.533.6+3.0
AA-3D v1.0 EloCode974954−2.9
Artificial Analysis Intelligence Index v5.0.1Indexes30.833.7+2.8
GPQA Diamond (Artificial Analysis avg@5)General QA95.392.8−2.4
Epoch Capabilities Index (ECI)Indexes159156−2.3
MKA (Artificial Analysis)Professional Work35.337.4+2.2
𝜏³-BankingAgents45.847.8+2.1
AA-PersonalityBench Organization indexChat & Writing73.371.4−1.9
AA-PersonalityBench Slop indexChat & Writing71.473.3+1.9
FrontierSWE V2Code19.617.8−1.8
MMMU ProMultimodal84.582.8−1.7
Medical Long Context Reasoning (MLCR-AA)Health & Medicine21.720.0−1.7
Terminal-Bench-Science 0.1Science10.511.9+1.4
AA-AnalystAgent (pass^5)Professional Work46.345.0−1.2
Artificial Analysis Progress Index v1Indexes320324+1.2
Terminal-Bench 2.1Code87.688.8+1.1
VIALS-AA (average of 3)Multimodal52.050.9−1.0
Humanity's Last Exam (Text-Only)General QA42.143.1+1.0
CritPt (avg@5)Science18.317.7−0.6
GDP.pdf (Artificial Analysis, All-pass)Professional Work22.822.80.0

Gemini 3.8 FlashvsQwen3.8 Max (0902)normalised score marginsData Nov 15, 2026Data: Artificial Analysis, AIMING Lab (UNC-Chapel Hill), BuseyBench and 8 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…