---
title: Large-Instance TSP on 16 Arm Cores: 0.692 % → 0.165 % Mean Gap, Still Behind LKH-3 | SpinForge
url: https://spin-forge.com/proof/s03
description: 0.165 %: mean gap, 8 held-out national TSPs (was 0.692 %). Each Improvement Confirmed on 8 Held-Out National Instances.
---

GARMO SOLVERS · S03

# Large-Instance TSP on 16 Arm Cores: 0.692 %  0.165 % Mean Gap, Still Behind LKH-3

Each Improvement Confirmed on 8 Held-Out National Instances.

0.165 % mean gap, 8 held-out national TSPs (was 0.692 %)

2/0/6 vs LKH-3 best config, held-out, 30 s

7/7 dev instances improved

0.020 % d18512 with the opt-in search mode (LKH-3: 0.049 %)

- Held-out 8: mean gap 0.692  0.224  0.165 %
- vs LKH-3 still 2/0/6 held-out (wins on ho14473 and ch71009)
- An opt-in search mode beats LKH-3 on brd14051, d15112 and d18512
[Request the full poster by email](https://spin-forge.com/contact?product=portfolio&brief=Please%20send%20the%20full%20poster%3A%20Large-Instance%20TSP%20on%2016%20Arm%20Cores%3A%200.692%20%25%20%E2%86%92%200.165%20%25%20Mean%20Gap%2C%20Still%20Behind%20LKH-3%20%28S03%29.%0A%0AOur%20organization%20and%20intended%20use%3A%0A)

## More garmo solvers benchmark results

- [Proof-Carrying Max-Cut on a GH200: 67 Biq Mac Optima Proved vs HiGHS's 49](https://spin-forge.com/proof/s01)
- [Weighted MaxSAT on a GH200: From 14/33/50 to 33/38/26 Against NuWLS-c-IBR](https://spin-forge.com/proof/s02)
- [A Native QCDCL Solver vs DepQBF: From 0 vs 4 to a 15–15 Tie, Then 15 vs 18 on Fresh Instances](https://spin-forge.com/proof/s04)
- [Vehicle Routing vs PyVRP: a C++ Solver Wins 37/26/21 at Equal Budget](https://spin-forge.com/proof/s05)
- [ISPD98 Bisection: Our Partitioner Beats KaHyPar and mt-KaHyPar at Equal Budget](https://spin-forge.com/proof/s06)
- [An Evidence-Driven Solver Router: Regret 0.046 vs 0.190 for the Best Single Backend](https://spin-forge.com/proof/s07)
- [A Verified MIP Race: HiGHS, SCIP and GPU Primal Heuristics Against cuOpt on MIPLIB 2017](https://spin-forge.com/proof/s08)
- [Constraint Programming in C++: 39 of 40 Schedules Proved Optimal vs CP-SAT's 20](https://spin-forge.com/proof/s09)
- [Certified GPU Numerics: Every Answer Carries a Checkable Certificate: and Where It Costs](https://spin-forge.com/proof/s10)
- [LABS on a GH200: Exponent 1.24 to the Proven Optimum, Beside the QAOA Exponents](https://spin-forge.com/proof/s11)
- [UNISOLVE 0.16.1: One Solver Platform, 13 Problem Domains, Every Answer Checked](https://spin-forge.com/proof/s12)
- [Four Certified Builds: Toroidal Proofs, Unroutability Certificates, CertBQP and a 235× Batch Service](https://spin-forge.com/proof/s13)

[All results, charts and papers](https://spin-forge.com/proof)

Results describe the stated inputs, hardware and conditions and are not a guarantee of results in your environment. See the [Terms of Use](https://spin-forge.com/terms).