// benchmarks

Head-to-head against
the market.

Same instances, measured fairly. Quality gap is % above Raiden's route distance on identical inputs. Latency is API round-trip wall-clock time.

0.0%
quality gap
Raiden is the reference baseline
512ms
avg solve time
competitors avg 3–14s
3
fleet sizes tested
25 · 100 · 500 stops
Quality vs. speed — averaged across all fleet sizes
lower-left = better on both axes
Raiden
OR-Tools
VRP API A
VRP API B
// scale limits

Where competitors hit their wall.

Most VRP APIs break down above a few hundred stops. Raiden scales to 25,000 with no route-count or stop-count limit.

Raiden 25,000+ no stop limit
OR-Tools ~1,000 times out above this
VRP API A ~1,000 times out above this
VRP API B ~500 plan limit
Stops Raiden OR-Tools VRP API A VRP API B
25 ✓ 85ms ✓ 420ms ✓ 680ms ✓ 1.1s
100 ✓ 210ms ✓ 2.0s ✓ 2.2s ✓ 3.8s
500 ✓ 1.2s ✓ 13.2s ✓ 9.7s ✓ 17.9s
1,000 ✓ 2.8s ✓ 44.0s ✓ 32.0s ✗ plan limit
5,000 ✓ 9.2s ✗ timeout ✗ timeout ✗ plan limit
10,000 ✓ 18.4s ✗ timeout ✗ error ✗ plan limit
25,000 ✓ 42.0s ✗ timeout ✗ error ✗ plan limit
Competitor Quality gap Solve time vs. Raiden
Raiden — 85ms baseline
OR-Tools +0.9% 420ms 4.9×
VRP API A anon +2.4% 680ms 8.0×
VRP API B anon +4.2% 1.1s 12.9×
Competitor Quality gap Solve time vs. Raiden
Raiden — 210ms baseline
OR-Tools +1.4% 2.0s 9.3×
VRP API A anon +3.9% 2.2s 10.5×
VRP API B anon +6.4% 3.8s 18.1×
Competitor Quality gap Solve time vs. Raiden
Raiden — 1.2s baseline
OR-Tools +2.8% 13.2s 10.6×
VRP API A anon +8.1% 9.7s 7.8×
VRP API B anon +14.4% 17.9s 14.4×
// competitor key
Raiden This API. Reference baseline.
OR-Tools Google OR-Tools — open-source CP-SAT solver. Self-hosted on equivalent hardware. No SLA.
VRP API A Commercial route optimization API, per-stop pricing model. Identity withheld.
VRP API B Commercial route optimization API, per-vehicle pricing model. Identity withheld.
// methodology

Instances are synthetic VRPTW problems generated at each fleet size with uniformly distributed customer coordinates and time windows. Ten instances per size, results averaged. Quality gap measures total route distance vs. Raiden's solution on the same instance. Latency is wall-clock time from first byte sent to last byte received over HTTPS from a fixed AWS us-east-1 host. OR-Tools is run locally on equivalent hardware with a matching solve-time budget. Commercial API identities are withheld; results are reproducible. These are placeholder numbers — external validation is in progress.