Atlas

Compare Models

Gemini 3.8 Flash vs MiniMax M3.1 Flash

Gemini 3.8 Flash leads by 0.8 capability points (≈95% interval −1.6 to +3.3): statistically tied · P(higher) ≈ 75%

Wins 28 of 39 shared benchmarks (11 lost)

Capability vs. relative cost
Gemini 3.8 FlashMiniMax M3.1 FlashClosed 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 FlashvsMiniMax M3.1 Flashcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins28 – 11 · 39 shared

Comparing matching setups only (56 benchmarks excluded) · suites counted once
← Gemini 3.8 FlashMiniMax M3.1 Flash →

Gemini 3.8 FlashvsMiniMax M3.1 Flashwins by categoryData Nov 15, 2026Data: Vals AI, Mercor, Scale AI and 16 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is MiniMax M3.1 Flash minus Gemini 3.8 Flash.

NoCoT-Bench NCKI (kspine v3)General QA12165.9−37.9
Toloka Arena (v3)Agents59.022.0−37.0
NoCoT-Bench NCRI 15.2General QA12668.4−31.5
ALE-BenchCode21821495−24.3
HiL-Bench (Human-in-Loop Benchmark)Agents41.565.7+24.2
BuseyBench SVGMultimodal6.64.2−24.1
DAXBench v2Code87.263.8−23.4
Visa Agentic Commerce BenchmarkAgents53.333.3−20.0
FrontierML ($200 wallet)Agents5.723.2+17.5
GBENCH One-shot Coding GScoreGames58.641.1−17.5
τ³-Banking (Mercor run, bm25_grep retrieval)Agents48.836.1−12.7
RSI Arena Stage 1 human ratingAgents9811002+10.1
Code Migration v1.1 (Vals Index v2.2 subset)Code38.248.2+10.0
Finance Agent v2.1Finance & Accounting65.858.8−6.9
Access to Justice BenchmarkLegal67.173.3+6.3
Humanity's Last Exam DiamondGeneral QA34.328.1−6.2
TaxCalcBench TY25Finance & Accounting6.00.0−6.0
LiveBench 2.0Agents67.373.1+5.8
DeepSWE v1.1Code73.768.1−5.6
Terminal-Bench 4.0Code19.725.3+5.6
Excel Modeling Benchmark v1.1Finance & Accounting72.967.8−5.0
Legal Research BenchLegal38.934.1−4.8
APEX-Agents 1.1Professional Work72.968.1−4.8
Vibe Code Bench v1.1Code78.773.9−4.7
HLAB-Vals-1.1Legal12.17.5−4.6
Vals Finance IndexIndexes60.555.9−4.6
Tax Agent BenchFinance & Accounting66.862.7−4.1
Professional Reasoning Benchmark - FinanceFinance & Accounting49.545.6−3.9
Professional Reasoning Benchmark - LegalLegal48.945.1−3.8
APEX-1 (Mercor, mean score)Professional Work68.865.0−3.8
CharXiv ReasoningMultimodal86.282.5−3.7
SAGE v1.1Math88.285.3−3.0
Vals Legal IndexIndexes69.867.1−2.7
Vals Index v2.2Indexes56.053.8−2.2
SWE-Bench Pro V2 — FullCode94.996.6+1.7
APEX-Accounting 1.1Finance & Accounting58.860.0+1.2
VoxelBench Text RatingGames17981810+0.9
Harvey LAB (Mercor run, undisclosed ~213-task cohort)Legal11.411.9+0.5
Design Arena Overall Code CategoriesCode12991297−0.3

Gemini 3.8 FlashvsMiniMax M3.1 Flashnormalised score marginsData Nov 15, 2026Data: Vals AI, Mercor, Scale AI and 16 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…