Warscape
Simulated matchup ยท a model, not a forecast

Azerbaijan vs ArmeniaMilitary comparison, and what happens if they fight

If Azerbaijan attacked Armenia today, with no allies joining on either side, this is what the WARSCAPE engine produced: Defender holds after 365 days.

How the two sides compare

Azerbaijan keeps 1.6 times as many soldiers under arms. Azerbaijan has 4.2 times the tanks. Azerbaijan has 1.9 times the combat aircraft. Azerbaijan spends 2.4 times as much on defence. Public, country-level figures from the data snapshot the run was pinned to (2026.09). See data sources for where each number comes from.

AzerbaijanArmenia
68,200Active soldiers42,900
300,000Reservists210,000
920Tanks221
600Artillery250
31Combat aircraft16
60Helicopters20
$3.9bnDefence budget$1.6bn
10.2mPopulation3.1m

What the war cost

Azerbaijan
attacked
4,340
soldiers killed or wounded
28
tanks destroyed
30
aircraft lost
0
land changed hands
45%
support at home, at the end
Armenia
defended
5,236
soldiers killed or wounded
6
tanks destroyed
8
aircraft lost
0
land changed hands
32%
support at home, at the end

What decided it

01
Logistics helped Armenia
Average front-line supply efficiency: 14% vs 50%.
02
Air situation helped Azerbaijan
Average share of air control for Azerbaijan: 63%.
03
Mobilisation helped Azerbaijan
Reservists mobilised: 271,593 vs 164,770.
04
Deployed combat power helped Azerbaijan
Average deployed combat power: 17 vs 10 model units.
05
Political cohesion helped Azerbaijan
Average cohesion: 58% vs 43%.
Think it would go differently?

This run assumed no allies, winter, full mobilization for the attacker and partial for the defender, and no nuclear weapons. Change any of that and watch it play out day by day on the map.

Other borders in this neighbourhood

All matchups โ†’
After 365 simulated days the conflict between Azerbaijan and Armenia ended with: Defender holds (duration limit reached). Aggregate personnel losses within the model: 4,340 (Azerbaijan) and 5,236 (Armenia). This page reports one seeded run of a model on public, aggregate data. It is not a prediction, and no operational information is used.

Model-generated simulation on public, country-level data. Not a prediction. No operational, classified or real-time information is used.