The MMU3 has been our workhorse for single-nozzle multi-material printing, and before it hands the baton to the upcoming INDX system, we wanted to give it one last significant upgrade. We usually design our hardware with some headroom, so it can be pushed further even with software updates. And the latest firmware update for MMU3 delivers a 20% speed improvement to filament changes, dropping from 52 seconds down to ~42 seconds per swap, saving up to 9 seconds. That might sound modest until you realize what it means across an entire print: total print times are slashed by 2-6 hours (or even more) depending on the model complexity!

And by the way, we can consider these numbers still rather conservative – we want to prioritize reliability across the entire user base. In case you would like to experiment, and your setup is flawless, it is theoretically possible, e.g., to push the loading speed up to 200mm/s and save even more time.

Why 9 seconds is huge

Multi-material prints rack up filament changes fast. In fact, on a complex 5-color model, like the Deadpool bust in the picture below, up to 60% of the total print time is spent entirely on non-printing tasks like load/unload sequences and wipe towers – and this is just the MMU3, which is already among the most efficient solutions on the market.

The example model uses a 0.2mm layer height, and it’s the well-known Deadpool bust remix by MacDonald Print Works.

A moderately complex five-color print might perform 400 (or more) filament swaps. By shaving 9 seconds off every single one of those, you are saving an hour on a single print. Scale up to a detailed model with 2,200 changes, and you’re looking at five and a half hours saved.

Nemo by Cipis

The math gets even better on production runs. If you’re printing a batch of 25 small items with 150 changes each, you save a significant amount of time. For anyone running a small business or cranking out prototypes, this is a huge timesaver.

With this optimization, the MMU3 combined with the CORE One (using a 0.2mm SPEED profile and a standard 0.4mm nozzle) quickly climbs the speed charts:

  • Print Speed: The MMU3 is approximately 40-50% faster on average with PLA compared to similar competitor designs (single nozzle multi-material) using default configurations.
  • PETG Performance: It is also roughly 18% faster on average with PETG compared to the competition’s default configurations.
  • Filament Waste: Thanks to our current implementation of ramming and stamping, the wipe tower remains contained and compact. As a result, the MMU3 wastes an average of 73% less material (in default config) than its competitors when printing PLA and PETG.

Printing the Benchbin is faster by 2 hours! Dropping from 23:41 to 21:48.

Note: We decided not to publish a table with a speed overview/improvements. This is mainly due to the fact that various setups might affect the overall time, e.g., when the MMU3 repeats a filament load, it adds to the total print time. We’ll be more than happy if you decide to share your experience and observations – even compare it to other printers!

Ready to jump in and test the new speeds? Let’s go through a quick checklist:

Before you start your first high-speed print, please ensure that the correct PTFE length for your setup is set in the printer’s menu via Settings -> Hardware -> MMU -> Front PTFE Length. When in doubt, see the MMU3 Handbook.

Please keep in mind that proper maintenance and correct setup have a large impact on the overall performance.

The filament swap time is “faster by up to 9 seconds” because it also depends on the length of the PTFE tube – some printers have a longer PTFE tube, which adds some hundreds of milliseconds to the loading process.

Derpy Tiger Painted by Tommy_Prusa3D (Remixed from Derpy Tiger by Heidi Makes)

The Lighthouse contains 3777 filament changes, which means saving 33,993 seconds or 9,5 hours!

And finally, a quick note on nozzles before you print: While it might be tempting to use a High Flow (HF/CHT) nozzle to speed things up further, we strongly advise sticking with the Standard (STD) nozzle for MMU prints. A high flow nozzle only offers a theoretical 2% speed improvement for multi-color prints, but standard nozzles are much better at ensuring a clean filament unload, less waste, and generally very fast prints in real scenarios. You can learn more here.

Technical Details: How We Did It

If we look closely at a filament change, the unload/load movements and the cooling moves on the wipe tower represent a massive 60% of the swap time for PLA and 70% for PETG. It was a bit like building a streamlined version of a sports car for track use – you need to shed as much weight as possible. So it’s a couple of grams here, a couple of grams there.

The speed gains come from optimizing these mechanical motion sequences.

Here are a couple of areas we focused on:

Parallel Idler Disengage
During filament loading, once the extruder takes over from the MMU, it no longer waits for the MMU to fully disengage the idler — a step that previously took about 2 seconds of dead time. Now the extruder pushes filament into the nozzle immediately while the MMU disengages in the background.

Predictive Idler Staging
During unloading, while the printer retracts filament from the nozzle, the MMU simultaneously moves the idler from its parked position to a ready stance — just short of engagement. Engaging from the “park” position normally takes over a second, but pre-staging virtually eliminates that wait.

Smoother Unload Ramp-Up
A refined pressure-relief step right before the unload sequence prevents extruder motor stalls, allowing us to safely increase retraction speed.

These are just some of the examples – we had to optimize more areas to reach the aforementioned 9 seconds, basically shaving hundreds of milliseconds wherever we could.

Every supported printer benefits immediately: MK4S, MK3.9, MK3.5, and CORE One/One+ users just need to flash new firmware(s) – one for the MMU3, and one for the printer itself. The correct firmware versions are 6.5.3 (printer) and 3.0.4 (MMU3).

Quality of Life and Reliability Tweaks

Speed isn’t everything; we also wanted to make the MMU3 less of a hassle to manage. By popular demand from our beta testers, we’ve added a “Preload All” feature. You can now preload all filaments of the same type (like all your PLA spools) at once, saving you from clicking through the menu for each individual slot.


Ripper Mask by Borderlands

We’ve also squashed a few bugs to improve overall stability and you can find all the details in full release notes.

CORE One L and MMU3: The Path Forward

We also had to make a decision regarding the CORE One L – it wasn’t an easy decision, but in the end, it makes the most sense. While we initially planned official MMU3 support, the imminent arrival of INDX shifted our priorities.

Instead of developing a dedicated MMU3 kit for the CORE One L, we’re providing adaptation guidelines for users who want to modify the CORE One+ MMU3 to work with the larger machine.

One caveat is that you will also need larger buffer plates – they need to be sourced separately from your local hardware store, etc., as they cannot be easily printed, and we provide STEP/DXF files to enable alternative manufacturing methods. Please check Printables and the article linked above for more info.

A goodbye gift (except it isn’t goodbye)

The MMU3 remains an exceptional single-nozzle multi-material system. Its minimal waste output, reliable tip forming, and now significantly faster operation continue to make it a solid choice for anyone looking for five-color printing.

Model 7 Buggy by 3D Sets

This firmware update is our thank-you to the community that’s pushed the MMU3 to its limits. Whether you’re printing beautiful visual models, functional prototypes with dissolvable supports, or production models, you’ll immediately notice the extra speed in every print.​

The INDX will bring even better speeds and up to 8 toolheads, but the MMU3 still remains a great choice, and we’ll keep offering it as long as there is interest in it.

Your MMU3 just got faster – go make something colorful with all that extra time! 😉