Gephi is an open-source desktop application for exploring and visualising networks, letting analysts import graph data, run layout and centrality metrics, then export publication-ready network diagrams.
Gephi is an open-source network analysis and visualisation application built for people who need to make sense of relationships rather than rows. You import a graph from GEXF, GraphML, CSV or Pajek files, apply a layout algorithm to position thousands of nodes, then colour and size them by metrics such as degree, centrality or community. Filters in Gephi narrow the picture while you work, and preview mode renders a clean export for reports. It is exploratory software rather than a charting library: the loop of layout, measure, filter and inspect is where insight actually appears.
Gephi earns its place once a spreadsheet stops being readable. The engine is built in Java and tuned for interactive rendering, so panning, zooming and recolouring a graph of tens of thousands of edges stays responsive on ordinary hardware. Because layouts are recalculated on demand rather than baked into the file, you can compare ForceAtlas2 against a circular arrangement and keep whichever reveals the structure. Networks are visual by nature, which makes accessible alternatives worth planning: exporting node and edge tables alongside the diagram helps colleagues who read results through a screen reader such as NVDA Beta. Community-built plugins extend layouts, metrics and export formats without touching the core.
The practical gain is speed of understanding. A table of forty thousand transactions tells you nothing at a glance; the same data laid out in Gephi exposes the handful of brokers, outliers and tight clusters that actually matter, usually within minutes of import. That matters in fraud review, citation mapping, supply-chain risk and social research, where the question is nearly always who connects to whom rather than how many records exist. Gephi keeps the measurement honest as well: betweenness, modularity and clustering coefficients are computed on the same view you are looking at, so a presentable picture and a defensible number come from one workspace. Because the project is open source, methods can be checked, extended and cited in academic work, and results stay reproducible when you share the project file instead of a screenshot.
Key capabilities and highlights.
ForceAtlas2, Fruchterman-Reingold, OpenOrd and circular layouts reposition nodes according to their connections, turning a dense hairball into visible clusters. Each algorithm exposes its own repulsion and gravity settings, so you can decide how tightly communities collapse or how far isolated nodes drift away.
Degree, weighted degree, betweenness, closeness, eccentricity, clustering coefficient, PageRank and modularity run directly on the loaded graph. Results arrive as sortable columns you can colour by or filter with, so a ranking table immediately becomes a visual hierarchy instead of a separate spreadsheet exercise.
Modularity-based clustering partitions a graph into communities and assigns every node a class, which can then be coloured automatically. For analysts mapping factions, audiences or dependency groups, this surfaces structure you would otherwise have to guess at, and the modularity score indicates whether the partition deserves trust.
A spreadsheet view of nodes and edges lets you add, merge, delete and compute columns without leaving the application. Joining an attribute CSV to existing nodes is straightforward, which is how most real projects attach labels, categories or weights to a topology that arrived bare.
Filters apply to the canvas live, so you can keep only nodes above a degree threshold, inside a date range or belonging to one partition and watch the network restructure instantly. Adding and removing filters during exploration is quicker than editing the source file, and the filtered view exports on its own.
Graphs carrying timestamps can be scrubbed along a timeline, with nodes and edges appearing or fading as their activation window opens. Growth patterns become legible that way: you watch a cluster detach or a hub emerge instead of comparing two static snapshots and guessing what happened between them.
Preview mode renders the graph with curved edges, labels, custom fonts and a chosen background, then exports to PNG, PDF or SVG at print resolution. Because preview is a separate rendering pass from the working canvas, you can polish a figure for a paper without disturbing the exploratory view underneath.
The plugin catalogue adds layouts, metrics, generators, filters and importers maintained outside the core release, installed from inside the application. A scripting console handles repetitive preparation steps, which helps when the same cleaning routine has to run across dozens of network files with identical structure.
Browse historical releases and previous builds.
Common questions about Gephi.
Yes. Gephi is open-source software released under the GNU General Public License, so there is no licence fee, seat limit or trial period. It can be used commercially, modified and redistributed under the same terms, and the source code is publicly available for inspection.
GEXF, GraphML, GML, Pajek NET, DL and CSV or Excel edge tables are all supported, and some plugins add further formats. CSV import includes a wizard for declaring which columns hold sources, targets, labels and attributes before the graph is built.
Layout and metrics are limited by available memory rather than a fixed node cap, so the ceiling depends on the machine and the memory allocated to the application. Interactive work is comfortable in the tens of thousands of nodes; beyond that, run statistics first and reserve visual layout for filtered subgraphs.
No. The standard workflow of import, layout, measure, filter and export happens entirely through menus and panels. Coding only becomes relevant if you want to script repetitive preparation steps or write a plugin of your own.
Yes. Preview mode exports PNG, PDF and SVG at chosen dimensions, with control over labels, edge curvature and background. SVG is usually the better choice for print because it scales without pixelation when a figure is resized in layout software.
It does. GEXF files can carry time intervals for nodes and edges, and the timeline control replays that data so the structure can be watched evolving rather than compared as separate snapshots.
Because it is written in Java, it runs on Windows, macOS and Linux with broadly consistent interface behaviour. A project file created on one platform opens on another provided the same plugins are installed.
Yes. The application runs locally and analysis needs no internet connection. A connection is only required when browsing and installing plugins from the catalogue or following documentation links from inside the application.
Gephi is an open-source desktop application for exploring and visualising networks, letting analysts import graph data, run layout and centrality metrics, then export publication-ready network diagrams.
NVDA RC is the release candidate build of the free NVDA screen reader for Windows, published ahead of each stable version so testers can try new speech, braille, and browsing changes early.
NVDA is a free, open-source screen reader for Windows that speaks what is on screen and follows the keyboard, giving blind and low-vision users access to applications, documents, and the web without a paid licence.
Syncthing is an open-source program that keeps folders identical across your own devices in real time, with no cloud service in the middle. Files move directly over encrypted connections, and only devices you explicitly approve are allowed to connect.
qBittorrent is an open-source BitTorrent client for downloading and sharing authorized files through torrent or magnet links, with queue management, bandwidth controls, content selection, searching, RSS automation, and remote access.
Comments