Asahi Kasei KARAKSA Review: CNF Is a Whole New Option for Fiber-Reinforced Nylon Filament

Read in other languages

Hey, it’s YuTR0N.

[Note for readers outside Japan] This is the English version of a review I originally published in Japanese. KARAKSA is an Asahi Kasei product and, as of writing, it’s sold through Japanese shops — purchase links and the contact route for overseas inquiries are at the bottom of this article.

This time Asahi Kasei sent me a genuinely interesting filament. Yes — that Asahi Kasei, the company behind the cling wrap that’s probably sitting in your kitchen right now. They also joined JRRF2025 last year as a sponsor.

“A fiber-reinforced filament that’s happy with a brass nozzle, and happy with a 0.2mm nozzle.”

That sounds like a fantasy, but that’s exactly what the PA-CNF family — DX and BX — claims. You may have already seen sample parts at trade shows. The filament is called KARAKSA.

Most of what I print with day to day is, honestly, the easy stuff, and the mental image I’ve always had is “filament must be bone dry, no exceptions!” — but KARAKSA turned out to have a few surprising sides to it. (More on that later.)

Writing the H2S review taught me a lot about engineering materials, and this project pushed that knowledge further again.

Related (Japanese): Bambu Lab H2S Combo hands-on review

One more thing worth mentioning given how noisy the supply chain has been lately: with KARAKSA, the cellulose nanofiber (CNF) production, the compounding with resin, and the extrusion into filament are all done domestically in Japan. That’s a different level of reassurance compared to a filament that’s made entirely overseas and imported.

*This article contains affiliate links.

So What Is CNF? It’s Not CF, and It’s Not GF

Fiber-reinforced filaments come in a lot of flavors, but what you can buy off the shelf on Amazon is mostly CF (chopped carbon fiber) or GF (glass fiber). This is a different animal: CNF — cellulose nanofiber.

Because the fibers are so fine, you don’t get as much of the “layer lines disappear into the texture” effect that GF and CF surfaces give you. On the flip side, you also don’t get that classic CF experience where you sand a printed part and end up with tiny fibers stuck in your fingers.

It also has strong thixotropic behavior.

In practical terms: it handles overhangs well, and it prints cleanly with crisp cutoffs instead of drooping and stringing everywhere. That’s probably the most useful way to translate the spec sheet.

I casually dropped brass nozzle and prints fine at 0.2mm at the top of this article, so here’s the why:

CNF is 10–100 nanometers in diameter. Carbon fiber is around 7 micrometers.

If your unit conversion just got fuzzy, don’t worry, let’s refresh it together.

1 micrometer [μm] = 1,000 nanometers [nm]

So you can see just how fine these fibers are. That’s why wear stays low even on a brass nozzle. Now, most recent machines ship with hardened steel or stainless nozzles anyway, so you might think the benefit disappears — but no, it absolutely doesn’t. Regular fiber-reinforced filaments can clog even a 0.4mm nozzle, and anything that reduces how often you have to swap parts is a real win. On top of that, less abrasion means the nozzle diameter barely drifts, so you can print at consistent quality over long periods.

And with all of that going for it, the heat deflection temperature at 0.45 MPa is 170–190°C.

Sliding performance is excellent too — in a reciprocating sliding test under a 2 kgf load, it posts a dramatically lower COF (coefficient of friction) than the alternatives.

Reciprocating sliding test method
Test method: from Asahi Kasei’s materials
Sliding test results
Test results: from Asahi Kasei’s materials

Here’s part of the actual test data.

ABS PA6/CF KARAKSA™ F
PA/CNF
Specific wear rate
mm³/(N·m)
ABS wear scar PA6/CF wear scar KARAKSA F PA/CNF wear scar
Specific wear rate
mm³/(N·m)
1.98×10−3 1.08×10−4 5.62×10−5
ABS
ABS wear scar
Specific wear rate mm³/(N·m)
1.98×10−3
PA6/CF
PA6/CF wear scar
Specific wear rate mm³/(N·m)
1.08×10−4
KARAKSA™ F PA/CNF
KARAKSA F PA/CNF wear scar
Specific wear rate mm³/(N·m)
5.62×10−5

The First Hurdle

Let’s start with prepping a fresh spool.

The recommended drying conditions are:

  • Hot-air dryer: 100°C × 7h, or 120°C × 2–4h
  • Vacuum dryer: 80°C × 48–72h

That narrows things down fast. Unless you use one at work, a vacuum dryer isn’t exactly a household item… which leaves the hot-air option. But a consumer filament dryer that hits 120°C? Good luck.

There’s no easy-to-buy unit in that range. The closest is the EIBOS E2, and even that tops out at 110°C — just barely short — plus it takes up a lot of space. So for now I went with what I already had on the shelf: the AMS HT (85°C) [US/UK/EU/AU store], the second-generation EIBOS Polyphemus (80°C), and the Creality SpacePi X4 (85°C).

In practice, here’s what I found: if you open a fresh spool and start printing right away, and then store it sealed with desiccant whenever it’s not in use, running roughly 8 hours of drying at the dryer’s default setting before a print is enough to work with it. (I haven’t tested it through Japan’s rainy season or peak summer humidity yet.)

Print Settings and Prep

Testing was done mainly on the Bambu Lab H2S and P2S, the QIDI Q2, and the Snapmaker U1.

I confirmed on the P2S that CNF is fine at 0.2mm, so on the H2S I went with a 0.4mm HF nozzle — not that it’s required. (You don’t get much benefit from an HF nozzle here, so a standard one is perfectly fine.)

For the build plate, every machine ran a textured PEI plate (the stock plate) with unscented Cape 3D Extra Keep spray applied.

This 300g × 2 pack is the one I recommend (Japan):

If you’re buying it yourself that’s fine, but if the company is paying… buy it by the case! It’s a consumable! It’s cheap!

If you really can’t get that specific spray, a plain PVP glue stick works too. Gluing threads onto a lid with your finger and building up something fabric-like… okay, I’m getting nostalgic about middle school art class. Moving on.

Multi-Material Units Are Fine (AMS / QIDI BOX)

Bambu Lab’s AMS 2 Pro and AMS HT, QIDI’s QIDI BOX, and the Snapmaker U1 all handled it normally — no bypass path needed. Even if you’re not doing multi-material or multi-color work, being able to feed straight through these units makes life much easier. That said, leaving any filament loaded indefinitely is a risk, so unload it if you’re not going to print for a while.

On the P2S You Have to Trick the Machine

If you only ever print PLA you’ll probably never run into this, but recent-generation Bambu Lab machines love to advertise themselves as engineering-filament capable while giving you almost no options. At 0.4mm the profile list is fine; at 0.2mm there’s basically nothing. PC was there, though! So with the very simple logic of “it’s a high-temperature type, that’ll do,” I built my filament profile on top of the PC profile.

Bambu Studio filament profile based on PC

I’m suppressing the urge to shout “wait, doesn’t this mean plenty of other materials would print fine at 0.2mm too?!”

KARAKSA BX profile settings

For DX I used the same profile with only the temperatures changed.

KARAKSA DX profile settings

I was on a textured PEI plate, so I left the other temperatures alone. The flow rate (max volumetric speed) of 4mm3/s is a value I tested myself.

On the machine side I also forced the material type to PC.

At 0.2mm, always turn flow calibration off, and run calibration manually the first time. (Honestly, it’s hard to judge the result, so just printing with flow calibration off might be the easier starting point.)

As for the H2S — Generic now includes PA-CF, so basing your profile on that works fine. Back when the machine first launched I remember a lot of things being missing.

My H2S was provided by Sunstella (Japan): https://psych0h3ad.link/SBBLH2S

Distributors sometimes have stock when Bambu Lab’s own store doesn’t, so it’s worth checking both — including the official store (JP/US/UK/EU/AU). A P2S review is coming soon too.

The QIDI Q2 Needs Some Fiddling Too

Modern machines are all a little awkward in their own way, aren’t they.

On the QIDI BOX side you can set the material to PA-CF, but when you sync from the slicer, it won’t accept it.

The QIDI BOX is set to PA-CF, and sure enough it shows up when you press sync… it really is PA-CF, but:

QIDI BOX material set to PA-CF

Where did it go! That’s not okay!!

Slicer failing to accept the synced material

So you need to create a profile and switch it manually.

I read through QIDI’s PA12-CF filament settings and used those as the base, then adjusted from there.

QIDI slicer filament selection

You have to specify it like this every time. I expect this will get fixed in a future update. (I’ve filed a request.)

QIDI filament profile settings

I’d also recommend dropping sparse infill speed down to 100.

Sparse infill speed setting

Gap infill could probably come down as well, but this is the setting I actually used.

Gap infill speed setting

The Q2 quietly offers a large build volume at a restrained price, and it’s a machine I’d strongly recommend to anyone working with engineering materials. You can grab one here.

Related (Japanese): QIDI Q2 Combo hands-on review

Actually Printing With It

The drying requirements are demanding, but I never hit a case where a print simply failed — testing went smoothly throughout, which was a relief.

The one thing that puzzled me: right after a print finished — especially at 0.2mm — applying force to the part would make it crack apart alarmingly easily. I asked Asahi Kasei about it and was told that moisture absorption is the key. After that, no more problems.

Does It Warp on Large Parts?

Speed settings: on the H2S with the 0.4mm HF nozzle I didn’t really change the defaults, and instead capped things through the flow rate.

https://www.printables.com/model/1468403-large-object-print-test

Result: finished with no warping! I ran this one in DX, which specifically claims better warp resistance.

*The photo shows the part after destructive testing.

Large object print test after destructive testing

On a part this size, at peak flow you’re printing at roughly 200mm/s.

That said, faster isn’t automatically better. Breaking this large part apart, some of the failure clearly ran along the layer lines, so for real-world parts I’d lower the speed cap.

Broken large test part showing layer-line failure

I also ran a model on the Snapmaker U1 — a multi-color torture test with corners that love to warp — and it printed without drama. The U1 ships with a stainless steel nozzle, so being able to just use it as-is is reassuring.

https://www.printables.com/model/937776-the-benchbin-multi-color-3d-printing-torture-test

BenchBin torture test printed with KARAKSA

Even the lid section, where contact area gets minimal, stayed put and finished cleanly.

BenchBin lid section printed with KARAKSA

Seriously Good Overhangs

Since strong overhang performance is one of its selling points, I printed the VORON Stealthburner main body:

https://github.com/VoronDesign/Voron-Stealthburner/blob/main/STLs/Stealthburner/%5Ba%5D_stealthburner_main_body.stl

3DXTECH Obsidian™ PA6-CF

3DXTECH Obsidian PA6-CF overhang result

eSUN ePA-CF

eSUN ePA-CF overhang result

Bambu Lab PA6-CF

PRINSFIL PETG-GF

Bambu Lab PA6-GF

Bambu Lab PA6-GF overhang result

KARAKSA DX

KARAKSA DX overhang result

Here’s PLA vs KARAKSA DX on the underside overhang of that bin’s protrusion from the Snapmaker print earlier.
It was hard to see in photos, so I colored it in with a Gundam marker. PLA is a mess, while DX printed the overhang surface cleanly.

A Stage for My Digital Microscope

For some reason I felt like a university student again and went down a rabbit hole designing double helical gears from scratch.

Digital microscope stage with double helical gears

Ever since I watched a YouTube video about gear designs that suit 3D printing, I’ve made a habit of building double helical gears whenever I need gears. It’s a hassle, but still.

I didn’t want to make a screw for locking the vertical position, so the design leans on tight tolerances instead.

My first attempt used 0.4mm-nozzle instincts — loose clearances — and printed in things like Bambu Lab PA6-CF, which came out gritty. I considered loosening it further, but if a 0.2mm nozzle is genuinely usable, why not use it?

Between the tighter print tolerances and the different surface characteristics, the motion turned out remarkably smooth. The base and the rack gear are DX; everything else is BX.

Accuracy You Can’t Reach With a 0.4mm Nozzle

To be clear, I did not calibrate inner and outer diameters here — the point was to see how much changes from nozzle diameter alone.

I printed the motor mount section of this drone frame.

https://www.printables.com/model/1216454-project-synthara-generative-design-fpv-drone-frame

KARAKSA BX printed with a 0.4mm nozzle

“I printed it, but something’s off about the fit?” Sure, you can dial it in with settings, or change the model itself — but having “change the nozzle diameter” left on the table as a viable option is exactly where KARAKSA shines.

KARAKSA BX printed with a 0.2mm nozzle

You could of course tune pressure advance and flow to fill things in neatly, but getting a clean top surface without agonizing over it is another KARAKSA trait.

Surface quality comparison
From Asahi Kasei’s materials

Advice for Using It in Practice

First and most important: after a print finishes, wait for moisture absorption! Yes — please wait for it to absorb moisture.

Resist the urge to pop the part off the plate and rip the supports away immediately. Leaving it on the plate for about a day is the safer play. I’m impatient, so I took mine to the kitchen and put it under running water as a shortcut before removing supports. Depending on the geometry, it’s probably better to leave it on the plate and handle it gently so nothing bends. With a 0.4mm nozzle you may be able to get away with less care depending on the part, but at 0.2mm — take it seriously.

When I say “applying force,” what actually happened was: I lightly pried at a narrow support groove with a screwdriver, a crack snapped in, and the part split all the way around like a wheel of cheese.

Part broken before moisture absorption
Broken by applying force before moisture absorption

Break it after moisture absorption and you get this instead. The earlier part and this one are mirror-image halves of the same design.

Same print settings, completely different behavior. If you’re impatient like me, start with the running-water trick.

Who Is This Filament For?

People who’ve tried nylon-based fiber-reinforced materials and given up. People who can’t use CF, but found GF surfaces too rough and gritty to live with. People who couldn’t be bothered with constant nozzle swaps. And people whose printer at the time simply couldn’t handle it, so nylon never even entered the conversation.

Sure, you can make a lot of things in PLA or PETG — but depending on the application it gets hard fast, and plenty of PA fiber-reinforced filaments are genuinely difficult to work with. Filaments that won’t print on the stock profile the brand itself ships in the slicer are a great example. Bambu Lab’s PA lineup failed on default settings across the board for me; PA6-GF only worked once I dropped the bed to 80°C.

In that context, being able to just match the temperatures and dive in around 50–100mm/s is a rare kind of easy.

Speaking of sliding performance: guide parts, gears, drone frames, mechanisms you’d never have tried to print as a single piece before, and suspension parts all feel like good fits. Personally I mostly print printer parts, so — once I’ve finished the microscope stage I showed above — I’ll probably go back to printers and try building a sliding assembly entirely out of KARAKSA, just to see what happens.

Where to Buy and Further Reading

It’s currently available from Sunstella and SK Honpo — both Japanese retailers.

Sunstella Mall: https://psych0h3ad.link/KARAKSA

SK Honpo: https://psych0h3ad.link/sk-karaksa-f

KARAKSA PA CNF filament information:

http://www.asahi-kasei.co.jp/karaksa/

Asahi Kasei engineering plastics portal:

https://www.asahi-kasei-plastics.com/topics/tech-09/

Official X account: the KARAKSA brand now has its own X account. Follow it for filament news → @KARAKSA_ASAHI

If you have questions you’d like to put to Asahi Kasei directly — including inquiries from outside Japan — they’ve asked that you use the contact link on

this page (top right).

Asahi Kasei KARAKSA contact page

As for what’s next, I’m curious about color options and versions built on other base resins.

SHARE THIS ARTICLE
Post on X Reddit LinkedIn

Discover more from Psych0h3ad

Subscribe to get the latest posts sent to your email.

If you like this article, please
Follow !

Comments

To comment