Metal & ceramic sintering filaments
Published · Researched 2026-09-21
What if your filament printer could make real stainless steel? That's the pitch: metal-loaded filament — ~80–90% metal powder in a polymer binder — prints a "green part" you then debind and sinter in a furnace, burning out the binder and fusing the powder into dense metal. Stainless 316L, 17-4 PH, copper, bronze, titanium, tungsten, even ceramics. No six-figure metal printer required. Read the next sentence twice, because it's the whole review: the green part is not a metal part. Printing it is the easy 20%. Debinding, sintering, and hitting dimensions afterward is the actual process — and it's where every dream in this category either survives or dies.
What it is, in two sentences: sintering filaments are polymer–powder composites you print oversized on a normal FDM machine, then furnace-process into metal or ceramic parts. The market is two tracks: Virtual Foundry Filamet (PLA-binder, prints like PLA, hobby-accessible) and BASF Ultrafuse 316L/17-4 PH (catalytic debinding system, industrial partners, professional track).
What's genuinely good starts with the green part, and credit where it's due: Filamet genuinely prints like PLA. Nozzle 205–235 °C, bed 40–50 °C, a standard-ish profile — the MatterHackers listings confirm you can run this on a machine that yesterday printed a benchy. The powder loadings are serious: 85% for 316L, 88–90% for bronze, 78–82% for titanium, 91–93% for tungsten. That's real metal content, not metallic-effect glitter filament. The vendor shrinkage guidance is admirably honest about the physics: ~7% shrink gets you ~80–85% density; push to ~20% shrink and you reach the low 90s. Nobody's pretending you get full density for free. And the two-track market means there's an honest entry point: Filamet's workflow (print, debind, sinter via service or kiln) is genuinely accessible to a serious hobbyist, while Ultrafuse's catalytic system serves the professional who needs higher density and has a debind/sinter partner. For one-off metal parts with complex geometry — a custom bracket, a jewelry master, a prototype that needs to be actual steel — there is no cheaper path from CAD to metal.
What's actually wrong is the other 80%, and this pass surfaced a correction that inverts standard advice. Do NOT dry Filamet in a filament dryer. Virtual Foundry explicitly warns that dryers can adversely affect the metal-loaded filament. Every instinct from polymer printing says dry your spool; here that instinct damages the product. (BASF's drying guidance is a separate matter and wasn't re-verified — keep the two tracks distinct.) Then the nozzle: 85–93% metal powder through brass is a death sentence — 0.6 mm stainless or hardened nozzle is specified, and the starting flow is ~135% because the loading needs the extra push. Green parts are fragile before debinding in ways vendor copy underplays — handle them like the powder compacts they are. The density–shrinkage tradeoff is the category's central cruelty: want low-90s density? Accept ~20% shrinkage, with Z shrinking slightly more than XY thanks to gravity, and good luck holding a tolerance through that. The vendor's starting targets (~5% isostatic for copper/bronze, ~10% for steels) are workflow guidance, not a guarantee — dimensional control through sintering remains the hard, expensive skill, and debind/sinter service pricing and turnaround were unverifiable in this pass. Which raises the real question: the furnace. You either own a sintering kiln, pay a service bureau (quote first — this is where the "cheap metal parts" math often dies), or use Virtual Foundry's ecosystem. And the per-kilo economics never stop being brutal next to polymer filament — that's the category, not a complaint about any brand.
Who should skip it: anyone who thought "print metal parts" meant the printer does the metal part (it doesn't — the furnace does); anyone without a sintering plan before buying the spool (a drawer of green parts is just expensive PLA); anyone holding tight tolerances (design for ±shrinkage reality, or machine after sintering); anyone who won't buy the hardened 0.6 mm nozzle; and anyone tempted to "just try" the dryer on Filamet — read the warning above again. Also: if you need one metal part, price a CNC or DMLS service first. Sintering filament wins on geometry complexity and iteration, not on cost per part.
The two tracks deserve one more honest paragraph because they get conflated. Filamet's PLA-binder system means you can plausibly debind in a kiln and sinter with manageable equipment — it's the track for the serious hobbyist who already owns or will buy furnace gear. Ultrafuse's catalytic debinding is a different animal: the binder system is designed for industrial debind-and-sinter partners, the density ceiling is higher, and so is every cost around it. Neither is "better" — they're different sports. The mistake is buying Ultrafuse expecting Filamet's accessibility, or buying Filamet expecting Ultrafuse's density. Match the track to your furnace reality, not your ambition.
Where it genuinely wins: one-off complex geometry that machining can't touch — internal channels, lattice structures, organic shapes — in real steel, bronze, or copper. Jewelry masters, custom tooling inserts, prototype hardware, restoration parts for machines that haven't had spares in decades. The people getting good parts share a profile: they already think in terms of shrinkage compensation, they own or have access to a kiln, and they treat the first three sintering runs as tuition. If that's you, this is the most exciting filament category in the project. If it isn't, the tuition is brutal and the drawer of green parts is where projects go to die.
Pass 3 additions (2026-09-21): a sobering owner account from the Hobby-Machinist forums describes significant wasted time and material during process setup, severe shrinkage and porosity, and high furnace-temperature requirements — one owner account, not a verdict, but a useful counterweight to vendor marketing and to the "printing is solved" framing above. Retailer guidance corroborates the fragile-feedstock note (high metal loading makes filament brittle before melting: keep the feed path straight, a filament warmer may help relax spool curvature) and the 0.6 mm hardened-nozzle baseline, with alloy-specific exceptions. And a data-hygiene note: powder-loading figures vary across Virtual Foundry and reseller versions, so treat the 85/88–90/78–82/91–93% table above as one source's numbers, not gospel.
The social pulse: the Metal FDM 3D Printing Group (by The Virtual Foundry) exists on Facebook — one of the few topic groups in this whole project that actually matched a filament — but the big general groups returned nothing on sintering keywords, and Threads search was dead. YouTube carries the practitioner weight: kiln builds, sintering timelapses, density discussions. The shape of the community is telling — it's process nerds, not product reviewers. Nobody's unboxing Filamet for views; they're arguing about ramp rates and shrinkage compensation. The honest community verdict, pieced together: the printing is solved, the sintering is the craft, and the people getting good parts treat it as a furnace hobby that happens to start with a 3D printer, not the other way around.
Value call: this category doesn't have a per-kilo ladder that means anything, because the filament is maybe a third of the total cost — the rest is furnace time, debinding, and your learning curve. Filamet is the entry: cheaper spools, PLA-like printing, lower density ceiling, you manage the sintering. Ultrafuse 316L/17-4 PH is the professional track: catalytic debinding, industrial partners, higher density, professional money. Pick by your sintering plan, not by the spool price — the spool is the cheapest decision in the workflow. And before either: get a quote from a debind/sinter service, because that number decides whether this project happens at all.
The tradeoff: sintering filament is the cheapest path from CAD to real metal for complex one-offs, and the most misleadingly simple-looking filament in this project. The printer is the easy part. The furnace, the shrinkage, the density tradeoff, and the do-not-dry warning are the real material. Respect the second half and it works. Buy the spool first and figure out sintering later, and you own very expensive PLA.