The version 7 of Mercury has been released.
Dear Colleagues,
We are pleased to announce that version 7 of the Mercury tool has been released to the industry and academic community.
This new version fixes several bugs found in previous versions.
If you are a Mercury user, we strongly recommend getting the new version.
We hope you will enjoy it. If you have any suggestions or find any bugs, please report them to us.
Mercury 7 – click here to download
What’s new in Mercury?
Below, we describe the most important features and enhancements of version 7:
- AI-assisted model generation: Open Tools > LLM Model Generator, describe the system you want to model, and optionally supply a baseline model. You can review and edit the generated script, evaluate it locally with Mercury, and open it as a graphical model when the formalism is supported. The generator works with public providers and custom or local endpoints, including Ollama. Generation and evaluation are separate steps so you can check the model before using its results.
- Phase-Type absorption analysis for SPNs and CTMCs: Mercury can now report the mean and variability of the time to absorption, the probability of reaching each absorbing state, and the survival, cumulative probability, density, and hazard functions. For SPNs, the report identifies the corresponding absorbing markings. Optional symbolic computation produces closed-form expressions that can be viewed in LaTeX or exported to Mathematica.
- MTTA simulation with external functions: An SPN simulation can now call an external program, persistent process, or HTTP service to update timed-transition distributions as the simulation progresses. Input and output mappings can be tested before a run. The conventional MTTA workflow remains available without an external provider.
- Automatic warm-up detection in stationary SPN simulation: The minimum warm-up time is no longer treated as an automatic cutoff. Mercury checks trend and mean stability across all metrics before collecting steady-state batches. The chart highlights the three confirmation windows and marks the start of the stationary phase. You can zoom, pan, inspect individual experiments, and include the observations and calculations behind the decision in an optional Excel worksheet.
- Multi-metric and parallel experiments: You can evaluate multiple metrics in a parameter sweep, including transient CTMC and DTMC experiments and Fault Tree experiments. Independent experiment points can run in parallel for SPN, CTMC, DTMC, RBD, Fault Tree, and Event Tree models, subject to available CPU and memory. The results view includes per-experiment progress, logs, and charts, with CSV export and improved Excel export for simulations.
- Parallel transient SPN simulation and cancellation: Transient SPN replications can run in parallel. Long-running SPN analyses and analytical or simulation-based RBD and Fault Tree evaluations can be cancelled from the interface without reporting an interrupted run as successful.
- Event Tree modeling and scripting: Event labels support LaTeX, and definitions have editable identifiers that remain linked to their references when renamed. Use reach() to calculate the probability of reaching an event and def() to reference a definition. Branch probability validation, expression evaluation, experiments, and export to executable Mercury scripts have also been improved.
- RBD and Fault Tree evaluation. Experiments support compatible groups of metrics, including uptime and downtime. RBD reductions now preserve reliability and availability values when block parameters are edited, and equivalent parallel groups retain the same reliability regardless of reduction order.
- Scripting and a public Java API: Compatible metrics can be evaluated together in one solve call, while existing single-metric scripts remain supported. Java applications can now evaluate Mercury scripts without launching the graphical interface and receive structured results and diagnostics. Fixes also cover guards, dynamic arc multiplicities, and phase-type transition expansion.
- Numerical accuracy, compatibility, and usability: Transient CTMC and SPN analyses now avoid numerical underflow that could leave a curve with only its initial point. SPN metrics support natural logarithms, including LN{E{#P0}}. Other fixes cover DTMC transition expressions and transient charts, sensitivity analysis, and Coxian distribution fitting in Moment Matching. The updated manual is available under Help > Online Docs, and the About dialog provides a copyable BibTeX citation.
What is the Mercury tool used for?
Mercury enables a range of models to be created and evaluated for supporting performance and dependability evaluations, such as reliability block diagrams (RBDs), dynamic RBDs (DRBDs), fault trees (FTs), event trees (ETs), energy flow models (EFMs), stochastic Petri nets (SPNs), and continuous and discrete-time Markov chains (CTMCs and DTMCs).
The tool has been cited in more than 140 scientific publications and is used in over 18 countries around the world.
Important notes on upgrading
Java 8 (1.8) is the recommended version of Java for running Mercury 7. Java 8 introduces the Java Platform Module System (JPMS), a new architectural design to which the tool has not yet been ported. We deliver Mercury 7 with the appropriate Java SE runtime environment (JRE) for Windows and *NIX systems. We also release the tool without the JRE, which is suitable for other environments. If you want to download Mercury without the JRE, you can find the recommended JRE at https://www.oracle.com/at/
Enjoy Mercury 7!












