Plexus Mod Loader compared with Sinytra Connector and other bridges
There are two ways to get mods from different loaders into one game:
- A bridge on top of a loader. You install one loader, and a bridge mod makes it run another loader's mods. Sinytra Connector is one: it runs Fabric mods on NeoForge.
- A loader that hosts the others. Plexus Mod Loader is a loader of its own. It owns the game, runs Fabric and Quilt mods through its own support for them, runs Forge mods on Forge and FML themselves, and runs NeoForge mods on NeoForge and its FML themselves.
Both are good answers to different questions. This page lists the facts so you can pick the one that fits. The facts about Sinytra Connector come from its README and FAQ; check them there, as they may change.
Side by side
| Sinytra Connector | Plexus Mod Loader | |
|---|---|---|
| What it is | A translation and compatibility layer that runs Fabric mods on NeoForge | A mod loader that runs Fabric, Forge, NeoForge and Quilt mods, and its own Plexus Mods |
| Loader underneath | NeoForge | Plexus Mod Loader. Forge or NeoForge, each with its FML, runs on top of it when that loader's mods are installed |
| Mods it puts together | Fabric + NeoForge | Fabric + Quilt with Forge (MinecraftForge), or Fabric + Quilt with NeoForge. Forge + NeoForge in one game is not supported yet (POLY-COMPAT-221) |
| What you install | NeoForge, Forgified Fabric API, Launchpad, and Connector (per its README) | The Plexus Mod Loader installer. The first start with Forge or NeoForge mods downloads that loader's files |
| Fabric API | Forgified Fabric API, its modules rebuilt on NeoForge's own hooks | The regular Fabric API jar. On 26.2, none of its modules is left out next to Forge (1.2.5, 2026-10-28), and 10 are left out next to NeoForge (2026-10-27) (POLY-COMPAT-224); rebuilding those on NeoForge's hooks is planned |
| Minecraft versions | 26.2.x and 26.1.x supported, 1.21.1 as LTS with some bug fixes, 1.20.1 end of life (per its README) | 26.1, 26.1.1, 26.1.2, 26.2, 26.3 only |
| Stage | Beta (3.1.0-beta.3+26.2, checked 2026-10-19) | 1.4.0, a stable release |
| Where mixing breaks | Its FAQ names Fabric mixins aimed at methods NeoForge's patches changed, mods that keep separate code for each loader, and registries that NeoForge has already frozen when Fabric mods initialize | Fabric and Quilt mods whose mixins cannot work as written on the game Forge or NeoForge patched. Since 2026-10-25 such a mod is not loaded, and neither are the mods that need it (POLY-COMPAT-224, 114); the title screen counts them and the message says to report it to Plexus Mod Loader. On 26.2 that is Architectury, Balm and Lithium next to Forge (no Fabric API module; 1.2.5, 2026-10-28), and 10 Fabric API modules next to NeoForge. On 26.3, mixing loaders works since 1.2.5, with a few more mods left out (in 1.2.4 the game stopped at start). Mixins aimed at something the patch only moved are moved with it and still run (POLY-COMPAT-226). A mod with both a Fabric and a Forge or NeoForge build runs as its Forge or NeoForge build (POLY-COMPAT-139) |
Both tools face the same core problem: a Fabric mod expects vanilla code, and the other loader has patched that code. Neither can make every Fabric mixin land on patched code. Connector replaces Fabric API with a build made for NeoForge; Plexus Mod Loader does not have one yet, and leaves out the parts it cannot run instead of running them half-applied.
When to use which
- You play with NeoForge mods and want some Fabric mods: Sinytra Connector is a NeoForge mod built for this, with Forgified Fabric API. Plexus Mod Loader runs NeoForge mods and Fabric mods in one game too, but next to NeoForge it does not load 10 Fabric API modules yet (26.2, 2026-10-27), nor the Fabric mods that need those modules (below). If your Fabric mods need those parts of Fabric API, that is the case Connector is built for. Plexus Mod Loader with NeoForge was not measured against Connector, so the speed and memory results below do not cover that setup.
- You play with NeoForge mods only: Plexus Mod Loader runs them with NeoForge's own loader on 26.1, 26.1.2 and 26.2, also next to Plexus Mods.
- You want Fabric and Quilt mods together with Forge (MinecraftForge) or NeoForge mods on 26.1 to 26.3: Plexus Mod Loader does this for the mods whose mixins fit the host loader's patches; the others are left out with the reason (on 26.2: three Fabric mods next to Forge, 10 Fabric API modules next to NeoForge). On 26.3 this works since 1.2.5. Forge and NeoForge mods in one game are not supported. See Run Fabric and Forge mods together.
- You play on 1.21.1 or 1.20.1: Plexus Mod Loader does not support those versions.
- You make mods: Plexus Mod Loader also has its own Native API with no Minecraft types, for mods meant to keep working across game versions. See the introduction.
Sinytra Connector is a NeoForge mod that brings Fabric mods into NeoForge. Running it on Plexus Mod Loader's NeoForge has not been tested, so the two are not meant to be used together.
Measured comparison
On 2026-10-24 the same Fabric mods were run on Plexus Mod Loader and on Sinytra Connector on NeoForge, with vanilla as a third variant, on Minecraft 26.2's game test server (measurement 9). The ticks were then profiled and measured again (measurement 10). Plexus Mod Loader ran the Fabric mods without NeoForge; Plexus Mod Loader with NeoForge was not measured. A difference counts only when it is larger than the variation between runs.
| Plexus Mod Loader | NeoForge + Connector | Result | |
|---|---|---|---|
| Heap right after GC, largest | 173.1 MiB | 203.0 MiB | Plexus Mod Loader uses about 15% less (29.9 MiB). The runs did not overlap |
| Longest GC pause | 49.5 ms | 78.9 ms | Shorter on Plexus Mod Loader (44–60 ms per run, against 61–114 ms) |
| Server tick p50 | 8.47 ms | 8.48 ms | The same |
| Server tick p95 / p99 | 10.91 / 13.97 ms | 9.94 / 12.51 ms | No significant difference. Plexus Mod Loader did not win |
| Second start, to ready | 10.722 s | 10.232 s | No significant difference |
| First start | — | — | Not compared |
Medians of 5 runs from measurement 9. Measurement 10's 7 runs gave the same picture: heap 173.1 against 203.8 MiB, p95 10.73 against 10.07 ms, p99 14.18 against 13.07 ms.
- Memory. In every run Plexus Mod Loader stayed at 171–174 MiB and Connector at 201–207 MiB. The Connector side includes NeoForge itself, and the difference was not split between Connector and NeoForge.
- Tick tail (p95 and p99). Connector's p95 median was 0.98 ms lower in measurement 9 and 0.66 ms lower in measurement 10, and in an ordinary run it was lower than every Plexus Mod Loader run (4 of 5 runs, then 5 of 7). One slow Connector run in measurement 9 and two in measurement 10 make its standard deviation larger than the difference, so by the rule this is no significant difference. It is recorded as something for Plexus Mod Loader to improve.
- Why, from profiling the same runs. This is an analysis of the same runs, not a separate measurement. No cost that Plexus Mod Loader adds inside a tick was found (only 0.2% of the server thread's samples had a Plexus Mod Loader frame), so the loader code was not changed. The p95 difference sits in ticks 200–399, while the JIT is still warming up, and the NeoForge side warms up faster there. About half of the p99 difference comes from where a GC falls: Connector's first GC during ticks lands in the discarded first ticks, which leaves one GC fewer in its measured range. After tick 400, Plexus Mod Loader's p95 is the same as or lower than Connector's and vanilla's (measurement 9: 9.58 against 9.61 and 9.79 ms; measurement 10: 9.20 against 9.61 and 9.77 ms), within the standard deviation, so no advantage is claimed.
- Startup. Both were measured on a second start, with Connector's converted jars and Plexus Mod Loader's transform cache already in place. Connector was 0.49 s quicker to ready, within its own standard deviation (±0.62 s), so this is no significant difference. Plexus Mod Loader varied less between runs (±0.14 s). The first start was not compared: Plexus Mod Loader's cache was already warm from earlier test runs, so the conditions were not the same.
Mods only one side loaded
The comparison uses only the mods that loaded on both sides. In this test, on the 26.2 server, Connector 3.1.0-beta.3+26.2 could not load these Fabric mods, which Plexus Mod Loader loads:
| Mod | What happened with Connector |
|---|---|
| Architectury 21.1.12 | A mixin aimed at a method NeoForge patched failed its injection check, and the server stopped |
| Balm 26.2.0.9 | A mixin failed its injection check, and the server stopped |
| Lithium 0.25.3 | A mixin failed its injection check, and the server stopped |
| Puzzles Lib 26.2.5 | Connector's jar conversion stopped on one mixin class |
| Forge Config API Port 26.2.1 (Fabric build) | Its net.neoforged.fml.config classes clashed with NeoForge's own (LinkageError) |
| Zoomify 2.17.0 | NeoForge read the jar's neoforge.mods.toml and asked for a language provider that is not installed. It is a client mod |
LambDynamicLights is client-only, so Plexus Mod Loader leaves it out on a server, as Fabric Loader does (POLY-COMPAT-108); Connector loads it there. This is a compatibility difference in this test, not a speed result. Because Lithium was left out, its shorter ticks are not part of the comparison.
On NeoForge itself, Plexus Mod Loader does not run these five either. In a check that puts the 26.2 Fabric mod set on NeoForge with no NeoForge mods (corpusFabricOnNeoForge, the setup Connector is for, 2026-10-25), 37 Fabric mods load (96 before 2026-10-25, when the injections that did not fit were skipped) and 20 are left out with POLY-COMPAT-224: Fabric API modules, Architectury, Balm, Lithium and Puzzles Lib. Forge Config API Port's Fabric build is left out by the link check (POLY-COMPAT-111), because a Fabric API module it uses is not loaded. The game test server creates a world, passes GameTest and saves it. So, as with Connector, those five do not run; unlike with Connector, the game does not stop and each reason is given. Each mod's features were not checked. For a jar like Zoomify's, Plexus Mod Loader leaves the NeoForge side out (POLY-COMPAT-217) and runs the Fabric side when Fabric mods are present; this was checked with a test mod, not with Zoomify itself, which is a client mod. This is a compatibility result, not a speed result.
Scope
- One Minecraft version (26.2), one PC (Intel Core i9-14900HX, Windows 11, JDK 25.0.3, 4 GiB fixed heap), on the game test server only. Client frame time (FPS) was not measured, and neither was a dedicated server.
- 19 common Fabric mods: AppleSkin, Cloth Config, Collective, EntityCulling, Fabric Language Kotlin, FancyMenu, FerriteCore, GeckoLib, Jade, JEI, Konkrete, Melody, MezzConfig, NotEnoughAnimations, Placeholder API, 3D Skin Layers, Veinminer, Simple Voice Chat and YetAnotherConfigLib.
- Plexus Mod Loader 1.0.0 with Fabric API 0.161.0+26.2. NeoForge 26.2.0.93 with Connector 3.1.0-beta.3+26.2, Forgified Fabric API 0.161.0+26.2+4.0.2 and Launchpad 1.10.2+26.2, without the optional Connector Extras.
- Not measured: NeoForge alone (to separate NeoForge's share from Connector's), the 26.1 line, other hardware. The full list is under What is not measured.
What Plexus Mod Loader has checked
Only what Plexus Mod Loader's own tests show; these checks were not run with Sinytra Connector. For the head-to-head measurement, see Measured comparison.
| Loader | Status | What was checked |
|---|---|---|
| Fabric | Verified | On 26.2, 140 mods load, join a saved world on the client, play 200 ticks, save, and quit (138 real mods plus 2 test mods); 100 start a real dedicated server. 26.1 passes the same checks |
| Forge | Verified | Forge 65.1.5 on 26.2: 40 Forge mods on the client; 35 of 41 start a real dedicated server. Also checked on 26.1 and 26.1.2 |
| Fabric + Forge | Partial | Fabric mods whose mixins do not fit Forge's patches are not loaded (POLY-COMPAT-224): on 26.2 with 1.2.5, no Fabric API module, but Architectury, Balm and Lithium, and the mods that need them. With the rest, 88 mods of the other loaders create a world, pass GameTest, and save it on the game test server; 78 are left out (2026-10-28, 26.2). On 26.3 (since 1.2.5): 80 loaded, 86 left out |
| NeoForge | Verified | NeoForge 26.2.0.93 on 26.2: 54 NeoForge mods on the client; 37 start a real dedicated server (with a test harness mod). Also checked on 26.1 and 26.1.2. The mods left out are ones NeoForge's own server stops on or refuses too |
| Fabric + NeoForge | Partial | Fabric and Quilt mods whose mixins do not fit NeoForge's patches are not loaded (POLY-COMPAT-224): on 26.2 with 1.2.4, 10 Fabric API modules and the mods that need them. With the rest, 40 mods of the other loaders and 40 NeoForge mods create a world, pass GameTest, and save it on the game test server; 76 are left out (2026-10-27, 26.2). A network join into the play phase passed on 2026-10-25. 26.1, 26.1.2 and the client were not run again under this behaviour. On 26.3 (since 1.2.5): 40 foreign mods loaded, 76 left out, game test server and dedicated server. Not with Forge mods in the same game (POLY-COMPAT-221) |
| Quilt | Verified | On 26.1, 26.1.2 and 26.2, with every Quilt-only build on Modrinth for the version: on 26.2, 155 mods (88 Quilt) on the client and 133 on a real dedicated server. Next to Forge or NeoForge mods, the same limit as Fabric applies. The mods left out are ones real Quilt Loader cannot load or run either; mods that need QSL are left out (POLY-COMPAT-114 with 308) |
Each mod's features were not checked one by one. Details: Supported versions and loaders and Compatibility.
Other bridges
Other projects bridge loaders in other directions. Plexus Mod Loader has not tested them, so this page does not describe them.