Skip to content
Choose a tag to compare

Bleeding edge version


Created on 2022/05/18, 11:16 UTC by @awvwgk with commit 75575f4.
This is an automated distribution of the latest xtb version. This version is only minimally tested and may be unstable or even crash. Use with caution!

Choose a tag to compare

Many thanks to Christian Hölzer (@hoelzerC), Albert Katbashev (@Albkat), Jeroen Koopman (@JayTheDog), Hagen Neugebauer (@haneug), Felix Pultar (@pultar), Thomas Rose (@Thomas3R), Jordy Schifferstein (@Jordy-prog), Marcel Stahn (@MtoLStoN) for their contributions to this project and making this release possible.

New features

  • Implemented ORCA Hessian for thermo submodule (#504)
  • Add Oniom calculator (#573, #574, #577, #581, #578)
  • Use mctc library for geometry input / output, (#591, #594, #592)
    support QCSchema JSON, QChem molecule files and FHI-aims inputs
  • Allow printing of GFN-FF topology lists (#545, #570)
  • Add parameters for GFN1(Si)-xTB (#550)
  • Added C API option to call gradient calculations on point charges (#584)
  • Writing GFN-FF adjacency to output (#497)
  • Implement script driver (#579)
  • Reparametrization of Pt in Toluene for ALPB. (#524)

Bug fixes

  • Fix issues with structure converter (#604)
  • Catch out-of-bounds indexing for constraints (#602)
  • Don't check for existence of gfnff_topo (#595)
  • Fix hessian calculations for single atoms (#587)
  • Make sure to not allow filling more electrons than orbitals available (#526)
  • Fix overflow of array for long dynamics (#554)
  • Fix formatting in bond order printout (#544)
  • Fix duplicated key entry in JSON dump (#537)
  • Wrong cube sizing (#527)
  • Initialize number of electrons in SP calculation (#525)
  • Revert calculation of GFN-FF topological charges (#520)
  • Bump C standard to C11 for using _Generic (#623)


  • Add reference for ALPB and GBSA model (#502)
  • Use unit testing framework (#517)
  • Refactor CMake build files (#518)
  • Remove dead-code (symparam, symtrans) (#522)
  • Update meson build options (#533)
  • Update handling of charges in GFN-FF (#558, #609)
  • Refactoring of external drivers (#568, #582)
  • Add hessian calculation to calculator type (#569)
  • Cleanup printout of spin (#576)
  • Use iso_c_binding for building signal handlers (#622)

New Contributors

  • Christian Hölzer (@hoelzerC) made their first contribution in #497
  • Jordy Schifferstein (@Jordy-prog) made their first contribution in #527
  • Albert Katbashev (@Albkat) made their first contribution in #574
  • Felix Pultar (@pultar) made their first contribution in #584

Full Changelog: v6.4.1...v6.5.0


source linux windows

Choose a tag to compare

Bugfix and maintenance release, some new features sneaked in here as well (COSMO solvation model).

Thanks to Marcel Stahn (@MtoLStoN), Sebastian Spicher (@sespic), Christoph Plett (@cplett), Cyrille Lavigne (@clavigne) and Miguel Steiner (@steinmig) for contributing to this version.

  • Bugfix: topological charges in GFN-FF are now calculated correctly (#462) 🪲
  • Bugfix: projector matrix is now setup correctly in xtb-thermo subprogram (#441) 🪲
  • Bugfix: more stable parallelisation of numerical differentation (#465) 🪲
  • Bugfix: caching logic in Turbomole runner is fixed (#448) 🪲
  • Bugfix: transcription errors in (unused) STO-NG tables fixed (#464) 🪲
  • consistent solvent names for hexane (#436)
  • threadsafe caching for GFN-FF neighbourlists and charges (#467)
  • hessian projection of fixed atoms works now for L-ANCopt (#433)
  • COSMO solvation model for xTB (#274, #450)
  • Turbomole runner will use aoforce instead of numerical differentiation (#470)
  • reduced memory footprint for integral evaluation (#473)
  • cleanup test names for meson build (#460)
  • spurious warning for --parallel option fixed (#457)
  • turn unconverged SCF into actual error (#485)
  • fix keyword documentation for ESP grid file (#489)

source linux

Choose a tag to compare

Feature release with new GPU acceleration, improved ALPB parameters, better optimizer settings and options, and general performance improvements for GFN1-xTB and GFN2-xTB.
Thanks to Sebastian Spicher (@sespic), Marcel Müller (@marcelmbn), Hagen Neugebauer (@haneug), Thomas Rose (@Thomas3R), Marcel Stahn (@MtoLStoN), Julius Stückrath (@jbstueckrath), Patrick Atkinson (@patrickatkinson), Cyrille Lavigne (@clavigne) and Igor Gerasimov (@FulgurIgor) for contributing to this version.

  • Bugfix: Basename detection failed for paths with path delimiter (#362) 🪲
  • Bugfix: FOD flag (--fod) did not set electronic temperature correctly (#382) 🪲
  • Bugfix: Sign error in unbiased hessian calculation (#388) 🪲
  • Bugfix: IR intensities were calculated incorrectly from dipole gradients (#395) 🪲
  • Bugfix: Fix out-of-bounds access for basis set in IPEA-xTB calculations (#399) 🪲
  • Bugfix: Commandline interface for sequential build was inconsistent with OpenMP build (#421) 🪲
  • Bugfix: Point charge embedding was not initialized correctly and skipped embedding calculations (#422) 🪲
  • Bugfix: Negative charges were detected as command line flags (#423) 🪲
  • Improved 2D to 3D conversion with GFN-FF (#351)
  • GPU version of xtb is available with Nvidia HPC SDK (#342, #352, #407, #420)
  • Backports of performance improvements from GPU routines to CPU implementations (#425)
  • Allow adjusting of spring exponents in distance constraints (#357)
  • Allow scaling of dispersion energies / gradients for the GFN-FF (#359)
  • Improved ALPB implicit solvation parameters (#372, #387)
  • Turbomole driver now attempts to cache results via gradient file (#377)
  • Adjusted optimization threshold for level “lax” for Turbomole DCOSMO-RS driver (#383)
  • Allow reading static RMSD bias from file (#390)
  • Use averaged convergence scheme in L-ANCopt and RF-ANCopt to deal with noisy energy surfaces (#394)
  • Refactored hessian implementation to stabilize build and allow higher compiler optimization (#417)

linux linux

Choose a tag to compare

Maintenance and bugfix release.
Thanks to @haneug, @MtoLStoN, @sespic, @pprcht, @Thomas3R, @susilehtola and @timostrunk for contributing to this version.

  • ALPB solvation model has internal parameters now (new default)
  • --alpb/--gbsa gas/vac will not enable the solvation model
  • MTD can now use a static ensemble for RMSD bias as well
  • MTD time constant can be adjusted from input
  • MOPAC interface can now read dipole moments
  • new --bhess runtype to get hessians on non-equilibrium structures
  • Allow adjusting the imag. cutoff for thermo ($thermo/imagthr=real)
  • Allow scaling of frequencies in thermo ($thermo/scale=real)
  • Allow scaling of complete hessian ($hess/scale=real)
  • --json flag to enable dumping xtbout.json
  • Bugfix: Several workarounds for the NVIDIA compilers (former PGI) 🪲
  • Bugfix: Workaround GCC crashing on empty environment variables 🪲
  • Bugfix: Fixed non-deterministic failure of GFN-FF with GCC compilers 🪲
  • Bugfix: Charges are now handled correctly in PDB case 🪲
  • Bugfix: MOPAC interface reads total energy now from HEAT_OF_FORMATION 🪲
  • Bugfix: Fix error in GFN-FF torsion potential (new GFN-FF 1.0.2 version) 🪲
  • Bugfix: SDF input did not work with total charge in GFN-FF 🪲
Choose a tag to compare

Maintenance and bugfix release. Many thanks to @pprcht, @susilehtola and @felixmusil for contributing to this release.

  • Bugfix: Reference occupations for GFN2-xTB were incorrect for d-block elements 🪲
  • Bugfix: CT file charges were not read correctly 🪲
  • Bugfix: The loader API functions for xTB parameters did ignore the optional parameter files 🪲
  • Bugfix: Wigner-Seitz cell generation was erroring on certain architectures 🪲
  • Bugfix: OMP parallelisation of Hessian was not working with solvation models 🪲
  • Bugfix: Turbomole gradient file is written with $gradient but should be $grad 🪲
  • updated metadynamics pathfinder
  • ALPB model is now available
  • Failed geometry optimizations are now rescued to perform a final diagnostic SP before stopping the xtb run
Choose a tag to compare

Maintenance and bugfix release.

  • Bugfix: Total charge was not used in GFN-FF 🪲
  • Bugfix: Error handling in GBSA was not returned correctly 🪲
  • Bugfix: Copy mode overwrite of coordinate input 🪲
  • Bugfix: Negative charge was recognized as command line flag 🪲
  • Bugfix: Scan mode was not working correctly 🪲
  • Bugfix: GBSA energy was added twice for GFN-FF 🪲
  • Bugfix: Fix parallelisation issues in build (binary includes acf1ac7) 🪲
  • Logic for MKL linking has been reworked
  • PGI Compiler is now supported for builds with meson
  • Support for GCC 10.1 added

Please check version header for the correct version: 6.3.1-r1!

Choose a tag to compare

Feature release with a novel method, new runtypes and more periodic boundary conditions. Many thanks to Sebastian Spicher (@sespic) for contributing to this release.

  • Bugfix: missing symmetry numbers for S4, S8, T, Th and O 🪲
  • Bugfix: D4-ATM gradient for GFN2-xTB was slightly off 🪲
  • new format for the parameter files for the xTB Hamiltonians
  • GFN-FF implemented (see DOI: 10.1002/anie.202004239)
  • 2D to 3D structure conversion for 2D connection tables in (sdf/mol)
  • improved defaults for optimizer (RF based for less than 500 atoms, L-BFGS based otherwise)
  • GFN1-xTB available under periodic boundary conditions
  • new identifier system for the input
  • new submodule: info, thermo, topo
  • first stable C-API version (1.0.0)
  • Deprecation: Python API has been migrated to
Choose a tag to compare

Maintenance and bugfix release. Thanks to Christoph Bannwarth (@cbannwarth) and Jan Unsleber (@nabbelbabbel) for contributing to this release.

  • Bugfix: JSON dump did not return correct version number 🪲
  • Bugfix: Sign error in printout for GFN1-xTB Mulliken charges 🪲
  • Bugfix: PDB charges were not written correctly in PDB output 🪲
  • Bugfix: SRB strain derivatives were wrong 🪲
  • Bugfix: EEQ returns correct energy for one atom systems 🪲
  • The error handling has been completely reworked
  • Improved implementation of CM5 charges
  • Analytical gradients for the Halogen-bond correction
  • Benzene is now available as synonym for toluene for GFN2/GBSA
  • Massively improved C-API
  • Gaussian external input and output formats
  • CMake support for building xtb
  • GCC support for building xtb with both meson and CMake

This release is API compatible to version 6.2.2

Addendum: otool_xtb bash script to fix incompatibility of this release with the Orca 4.2.1 interface

Choose a tag to compare

Maintenance, bugfix and hotfix release.

  • configuration scripts are back in xtb
  • pkg-config file is now available
  • DFTB+ genFormat is supported as geometry input
  • additional check to prevent calculations on invalid geometry input
  • RF-Optimizer will not read hessian file by default anymore (now really)
  • GFN0-xTB can now be correctly requested from the $gfn group
  • Thermodynamic functions will not attempted to be calculated for T=0K anymore
  • Bugfix: periodic GFN0-xTB used wrong i-i' D4 dispersion interactions

This release is API compatible to version 6.2.1