Coverage ranges from 14% to 93% between the first and the last vendor: these APIs are not playing in the same league. Functional coverage matters more than algorithmic performance, because a fast engine that cannot model your critical constraints will never make it into production.
Route optimization has become a decisive factor in many industries: last-mile delivery, field service, waste collection, retail distribution. In response, the market offers a long list of APIs that all seem, at first glance, to answer the same need. In practice, coverage across this ranking ranges from 14% to 93% between the first and the last vendor.
This guide compares eleven solutions on their ability to model operational reality: breadth of constraints, optimization objectives, flexibility and fit with each use case.
Benchmark methodology
This benchmark reviews the public documentation of 11 vendors against 149 constraints representative of real-world operations. The constraints fall into eleven families:
- Objectives & balancing
- Cost & fleet economics
- Route structure & limits
- Pickups, deliveries & returns
- Capacity & load compatibility
- Service time
- Assignment, continuity & access
- Shifts, breaks & working hours
- Travel time & road network
- Real time & re-optimization
- Time windows
The full constraint list, with definitions and raw data, is available in the complete benchmark. Pricing for the eleven solutions is covered in a dedicated comparison.
Where a vendor ships both a solver you integrate yourself and a hosted API, it is the hosted API offering that is evaluated. That is notably the case for Timefold, where we analyzed the "Pickup and Delivery Routing" API rather than the open source solver. A solver you program yourself can, by construction, model almost any constraint; comparing it to an off-the-shelf API would make no sense.
Why comparing API performance makes little sense
Unlike conventional benchmarks, this analysis does not focus on algorithmic performance, meaning computation speed and solution optimality. Here is why.
The simplified-problem trap
An objective performance comparison would require testing every engine on identical problems. But because the available constraints differ so much from one vendor to the next, you end up falling back on basic cases: minimizing distance with a few time windows and capacity limits. On problems that simple, every engine performs roughly the same. The gaps are marginal, simply because the problem is too easy.
What actually matters: operational realism
An algorithm that returns its best solution in seconds matters far less than an algorithm that can model your operation exactly as it runs.
Take waste collection. Disposal sites often have limited intake capacity: one vehicle at a time during unloading, which takes an hour. If the optimizer cannot model that constraint and sends ten trucks in the morning, the whole day's plan collapses. The operator will never run the tool in production, however good its calculations look on paper.
That is why this benchmark puts functional coverage ahead of raw performance.
Our selection of the 11 best route optimization APIs
Kardinal
93% coverage · 1st/11A French company founded in 2015, Kardinal specializes in complex route optimization. The solution serves a wide range of industries: parcel delivery, e-commerce, waste collection, bulk delivery, field interventions, urban courier work. Its R&D team, which includes several PhDs in applied mathematics, built a proprietary optimization engine designed for maximum flexibility.
Kardinal covers 139 of the 149 benchmark constraints, or 93%, with only 10 gaps and not a single partially covered constraint. It is the only solution in the panel to fully cover eight of the eleven constraint families: objectives, route structure, capacity, service time, shifts, travel time, real time and time windows. Its weakest family remains assignment and continuity (67%), where Solvice does just as well.
Its 10 gaps all concern specific needs that only one other vendor covers: six variants of time-of-day hourly cost or per-visit facility cost and three cases of synchronized interventions, at Solvice, plus anchoring stops to a pickup point, at HERE.
Solvice
74% coverage · 2nd/11A European optimization specialist, Solvice targets complex use cases with a deeply technical approach. It posts the best score after Kardinal, carried by solid coverage across almost every family: objectives (72%), route structure (72%), capacity (89%) and above all pickups and deliveries (90%, the best score in the panel outside Kardinal).
Solvice shares the per-site throughput limit with Kardinal. Above all, it is the only vendor in the panel, Kardinal included, to model hourly cost by time of day, cost per visit at a facility and synchronized interventions between two crews: very targeted needs, rare in the sector's rate structures and day-to-day operations.
Its limit: no lexicographic optimization. With a model that rich, balancing several objectives means setting a weight on each criterion, and the configuration quickly becomes a heavy machine to fine-tune.
Solvice also offers a customizable business dictionary and a good grasp of European regulations, but remains less well known than the long-standing leaders, with a smaller ecosystem.
LogisticsOS
66% coverage · 3rd/11LogisticsOS is a US company founded in 2020 in San Francisco and a Y Combinator alumnus. Its API is built around cost optimization: usage-based billing and integration with the OpenStreetMap, TomTom and HERE Maps providers.
It matches Kardinal on the cost & fleet economics family (75% each): flat rates and overtime thresholds, combined billing across own fleet and subcontractors, guaranteed minimum revenue per subcontracted resource. It also holds up well on route structure (70%) and real time (88%).
For any operation where the economic trade-off between own fleet and subcontracting drives the decision, and if you are not afraid of fine-tuning parameters endlessly, LogisticsOS is worth testing.
HERE
59% coverage · 4th/11HERE Technologies, the mapping vendor known until 2015 as Nokia HERE, offers a Tour Planning API v3: multi-vehicle optimization backed by its own real-time traffic data and dedicated truck profiles.
HERE stands out with full coverage of travel time and road network (100%, tied with Kardinal and NextBillion AI). Notably, it is, alongside Kardinal, the only solution in the panel to support lexicographic optimization, the method that ranks objectives strictly rather than weighting them into a single formula.
If you already run on the HERE mapping data ecosystem, its optimization API is a natural candidate to evaluate.
Google offers a route optimization API backed by the scale of its infrastructure and its worldwide mapping data. It benefits from the excellent Google Maps base and native real-time traffic integration.
The solution is decent on basic objectives and predictive traffic, but sits well behind Kardinal and LogisticsOS on cost (50% against 75%) and lags on real time: unlike Timefold, NextBillion AI and Solvice, Google does not cover live problem updates while a route is being executed.
Google's infrastructure does guarantee excellent service availability, which counts when you operate at scale.
NextBillion AI
51% coverage · 6th/11NextBillion AI offers a mapping and optimization platform with a customizable business dictionary and native support for incremental optimization. It is strong on travel time (100%, tied with Kardinal and HERE) and real time (100%, tied with Kardinal, Timefold and Solvice).
A fair trade-off between targeted advanced features and more modest overall coverage, to evaluate first if reacting to disruptions matters more to you than functional breadth.
GraphHopper
40% coverage · 7th/11Open source by origin, GraphHopper also offers a commercial route optimization API. That dual nature brings transparency and an active community, with flexible deployment options: cloud or on-premise.
Overall coverage stays modest, held back by gaps on the cost, capacity and assignment families. GraphHopper holds up better on travel time (92%) and shifts and breaks (71%).
A good fit for small-volume operations, rather than for broad coverage of business constraints.
Timefold
32% coverage · 8th/11Timefold is a Belgian vendor based in Ghent, founded in 2023 by a team from the OptaPlanner project, whose open source solver it took over and developed further. Its positioning targets field service and workforce management first: technician routes, team schedules.
The overall score hides high marks on specific families: 100% on real time & re-optimization, tied with Kardinal, NextBillion AI and Solvice. Timefold's strength is concentrated in a few families rather than spread across the whole scope.
Relevant first for scheduling or field service cases through the hosted API, or for teams ready to invest in custom development around the open source solver.
Verso
29% coverage · 9th/11An entry-level API, Verso bets on simplicity and fast integration for standard needs. The API is deliberately lean, with few options, which makes it easy for technical teams to pick up.
One detail does surprise: Verso reaches 100% on the service time family, tied with Kardinal, Solvice, HERE and LogisticsOS. An island of completeness in otherwise limited coverage.
Suited to small fleets without sophisticated constraints, with the risk of having to switch solutions as operations grow more complex.
Routific
23% coverage · 10th/11In the same bracket as Verso, Routific offers a simple API aimed at small fleets, paired with an intuitive web app. It stands out for plug-and-play integration and a good reputation for responsive customer support.
Routific clearly targets small and mid-sized businesses with standard needs: pleasant interface, quick integration, but designed for a niche market.
Mapbox
14% coverage · 11th/11Mapbox, a reference in mapping solutions, offers an optimization API alongside its core products. It benefits from excellent geographic data and solid predictive traffic handling, naturally integrated with the rest of the Mapbox suite.
Optimization is clearly not its original business: functional coverage is the lowest in the panel. The API works more as a visualization add-on than as a genuine engine for complex optimization.
Ranking across the 149 constraints
| Rank | Solution | Coverage | Positioning |
|---|---|---|---|
| 01 | Kardinal | 93% | Generalist, complex constraints |
| 02 | Solvice | 74% | Generalist, complex use cases |
| 03 | LogisticsOS | 66% | Fleet cost trade-offs |
| 04 | HERE | 59% | Mapping and traffic |
| 05 | 57% | Global infrastructure and data | |
| 06 | NextBillion AI | 51% | Responsiveness and incremental |
| 07 | GraphHopper | 40% | Open source, small volumes |
| 08 | Timefold | 32% | Field service, scheduling |
| 09 | Verso | 29% | Entry level |
| 10 | Routific | 23% | Small fleets |
| 11 | Mapbox | 14% | Mapping add-on |
Kardinal keeps a clear 19-point lead over its closest competitor. The rest of the ranking tightens into a dense pack between 51% and 74% (Solvice, LogisticsOS, HERE, Google, NextBillion AI), followed by a group of more specialized or entry-level solutions at 40% and below.
What stands out most is how varied the profiles are inside that second group. HERE and LogisticsOS both sit ahead of Google across these 149 constraints, the former thanks to its road network, the latter thanks to its cost modeling. Timefold, conversely, posts a modest overall score but reaches 100% on real time and re-optimization, the mark of a deliberately vertical positioning.
Coverage by constraint family
| Family | Kardinal | Solvice | LogisticsOS | HERE | NextBillion AI | GraphHopper | Timefold | Verso | Routific | Mapbox | |
|---|---|---|---|---|---|---|---|---|---|---|---|
Objectives & balancing 36 constraints | 100% | 72% | 64% | 68% | 63% | 50% | 42% | 40% | 22% | 42% | 1% |
Cost & fleet economics 24 constraints | 75% | 54% | 75% | 40% | 50% | 17% | 4% | 0% | 4% | 0% | 0% |
Route structure & limits 23 constraints | 100% | 72% | 70% | 48% | 43% | 57% | 48% | 4% | 43% | 4% | 22% |
Pickups, deliveries & returns 20 constraints | 95% | 90% | 45% | 70% | 65% | 70% | 60% | 70% | 35% | 35% | 35% |
Capacity & load compatibility 9 constraints | 100% | 89% | 56% | 44% | 56% | 67% | 44% | 44% | 11% | 11% | 11% |
Service time 9 constraints | 100% | 100% | 100% | 100% | 89% | 22% | 22% | 11% | 100% | 11% | 11% |
Assignment, continuity & access 9 constraints | 67% | 67% | 33% | 33% | 33% | 33% | 33% | 17% | 0% | 17% | 0% |
Shifts, breaks & working hours 7 constraints | 100% | 71% | 71% | 43% | 71% | 57% | 71% | 36% | 36% | 43% | 36% |
Travel time & road network 6 constraints | 100% | 58% | 83% | 100% | 50% | 100% | 92% | 67% | 25% | 42% | 25% |
Real time & re-optimization 4 constraints | 100% | 100% | 88% | 63% | 50% | 100% | 0% | 100% | 63% | 25% | 0% |
Time windows 2 constraints | 100% | 100% | 100% | 100% | 100% | 100% | 50% | 75% | 50% | 50% | 100% |
Assignment, continuity and access remain Kardinal's weakest family at 67%, and Solvice does exactly as well. Service time and time windows, on the other hand, are market standards: most advanced vendors cover them fully or nearly so, which makes them poor criteria for telling solutions apart.
The capabilities that genuinely set vendors apart
Lexicographic optimization, which ranks objectives strictly instead of weighting them, is among the rarest constraints in the benchmark: only Kardinal and HERE support it out of the eleven solutions tested. Without it, every criterion needs its own weight, and configuring a multi-objective model quickly turns into a sprawling setup to fine-tune. Ranking keeps optimization simple and readable.
Predictive traffic, in contrast, is widespread: Kardinal, NextBillion AI, Solvice, GraphHopper, Google, HERE and Mapbox all cover it. LogisticsOS covers it only in part, and Routific, Timefold and Verso not at all.
Real-time problem updates are fully covered by Kardinal, Timefold, NextBillion AI and Solvice, and partially by LogisticsOS, Verso and Routific. Google, GraphHopper and HERE do not cover them.
Which API for which sector?
Overall functional coverage does not tell the whole story. Two APIs with similar general coverage can have radically different strengths and gaps from one sector to the next. What you optimize weighs as much in the decision as how complex your operations are.
Waste collection
Limited intake capacity at disposal sites and simultaneity constraints remain the most critical points; Kardinal is the only vendor to model them.
Read the full analysis →
Oversized & heavy freight (charter, LTL/pallets)
This segment demands fine-grained modeling of variable costs across vehicle and load combinations, plus multi-depot optimization. LogisticsOS stands out thanks to its handling of combined own-fleet and subcontractor costs.
Read the full analysis →
Field service / technicians
In field service, handling skills and certifications remains the central constraint. Timefold, despite positioning itself on this sector, does not make the top three: its hosted API targets a narrower use than route optimization at large.
Read the full analysis →
Long distance / FTL
Full truckload transport stacks regulatory, technical and operational constraints. Solvice leads the challengers, but 26 points behind Kardinal: the widest gap of any sector.
Read the full analysis →
Parcel & express
Parcel delivery remains the most demanding sector in constraint density: tight slots, high stop density, daily disruptions.
Read the full analysis →
Retail last mile / distribution
LogisticsOS leads the challengers in this sector, carried by its modeling of combined costs across own and subcontracted routes.
Read the full analysis →
Cold chain / fresh products
Fresh product delivery combines temperature constraints, short time windows and specialized vehicle profiles. Solvice comes second: alongside Kardinal and NextBillion AI, it is one of only three solutions that configure temperature compartments.
Read the full analysis →How to choose and test a route optimization API
The real value of an optimizer is not measured on a generic benchmark: it is measured on your data, with your constraints. Whether you run 3 vehicles or 300. Here are the criteria that actually make the difference when choosing and integrating a solution:
At equivalent functional coverage, these criteria determine integration speed, cost predictability and your ability to iterate fast.
Conclusion: which route optimization API should you choose?
There is no single best API. What exists is a fit between your real constraints and what each solution can model.
Functional coverage matters more than algorithmic performance. Kardinal covers 93% of the 149 constraints analyzed, against 74% for Solvice, the closest solution. Among the leading solutions, it is also the most competitive on price, including for small and mid-size fleets, as our 2026 pricing comparison shows. You can check it in minutes on your own data, for free.
Create a free account and get your first optimized routes in minutes, on your real data, with no commitment.
FAQ
Which route optimization API is the best?
There is no single best API: the choice depends on your sector and on how much of your operation goes beyond the classic constraints. Kardinal is the most complete solution on the market with 93% coverage across 149 constraints, ahead of Solvice (74%), LogisticsOS (66%), HERE (59%) and Google (57%). Start from these leading solutions, then rule out the ones that miss your critical constraints: real-time re-optimization, site simultaneity, hazardous materials, temperature compartments and so on. The more constrained your operation, the more the shortlist narrows around Kardinal, which is also the most competitive on price among the leading solutions, including for small fleets.
What is the difference between Kardinal and Solvice?
Solvice leads the other challengers with 74% coverage, notably on pickups and deliveries (90%) and capacity (89%). Kardinal still keeps a 19-point lead overall and fully covers eight of the eleven constraint families analyzed, against three for Solvice (service time, real time and time windows, three families where most advanced vendors already reach 100%). Above all, Kardinal offers lexicographic optimization, which Solvice lacks: without it, every criterion needs its own weight, and configuring a rich model quickly becomes a heavy machine to fine-tune.
Can you test a route optimization API for free?
Yes. Kardinal offers a generous free tier, with no time limit and no sales conversation, to test the API on real data. GraphHopper also offers a self-hostable open source version, as does the Timefold solver in its Community Edition. Access conditions for the other solutions vary (trial credits, limited free tiers or a sales contact): see the pricing comparison for details.
Did you evaluate the Timefold solver or its hosted API?
The hosted API. Timefold ships an open source solver (Community Edition) that you self-host and program yourself, plus a ready-to-use hosted API called "Pickup and Delivery Routing" with a fixed data model. This benchmark compares APIs you can consume as they are, so the 32% score reflects the hosted API. The solver itself can model far more, but only through custom development in Java or Kotlin, which puts it in a different category.
What is lexicographic optimization in logistics?
It is a method that sets an absolute hierarchy between objectives instead of combining them into a weighted formula. It lets you express, for instance: first minimize the number of vehicles used, then, at equal vehicle count, minimize total distance. Without it, every criterion needs its own weight, and configuration quickly turns into a sprawling setup to fine-tune as objectives multiply. Ranking keeps multi-objective optimization simple and readable. Of the 11 solutions tested, only Kardinal and HERE offer it.
Which API should you choose for a fast technical integration?
Kardinal is designed so that a developer gets their first routes on the very first API call, with no complex configuration. The API exposes consistent conventions, clear and actionable errors, idiomatic Python and JS/TS SDKs, and full documentation with reproducible examples. The free tier lets you run an end-to-end integration entirely on your own.