Cartesian by Formas review (2026): honest verdict on the AI 3D modeling tool — NURBS precision, natural-language modeling, CAD exports, and who it fits.
Cartesian is one of the more serious AI 3D-modeling attempts yet — it aims at precise, editable NURBS solids you can export to Rhino, SketchUp, and STL rather than the loose polygon meshes most text-to-3D tools produce, which is exactly what architects and product designers need. As a modeling tool for that audience it looks promising. The honest caveats are maturity and scope: it hasn't launched yet (public preview opens September 18, 2026), it has no published pricing and unverified output claims, and it is a modeling tool, not a content or marketing tool — it makes geometry, not posts. Score it as an early, credible CAD-adjacent modeler, not a finished product.
Cartesian is an AI 3D-modeling tool from Formas (formas.ai), a design-software company with offices in Singapore and Tokyo that also builds Formas Studio, an AI rendering and visualization workspace. Cartesian is the separate, CAD-adjacent product: you describe, sketch, or drag in a reference, and it builds a 3D model. Its pitch — "say it, sketch it, drag it in" — is aimed squarely at architecture and product design rather than at game-asset or hobbyist 3D.
This review scores Cartesian for what it is: an AI modeler for designers who need precision. I run a content engine, so the disclosure is upfront — Kompozy is a generation and publishing tool, and it isn't a 3D modeler — and I've tried to judge Cartesian on its own terms. The thing that separates it from the crowded text-to-3D field is the claim of native NURBS solids and clean, editable topology with named, related elements for BIM-style work, plus exports to SketchUp (.SKP), Rhino (.3DM), and STL. If that holds up, it's a meaningfully different tool from the mesh-blob generators.
The honest constraints matter as much as the promise. Cartesian hasn't launched yet: its public preview opens September 18, 2026, with early access going first to Formas' founding-circle and studio members and a waitlist for everyone else. There's no public pricing disclosed and no independent benchmarks of its geometry quality. AutoCAD and IFC support are listed as planned rather than shipped. So the scores below reflect its stated positioning and the demo material, not verified production output — treat them as a directional read to confirm in the app, and expect the picture to change as Formas documents the product. Everything here reflects Cartesian's public state on 2026-09-15.
Cartesian is a natural-language and multimodal 3D modeler: you type an intent, sketch a form, or drag in a photo, plan, or scan, and it produces a 3D model. Formas positions it as building "with maximum precision" — generating native NURBS solids and exact geometry rather than polygon-mesh approximations, with clean topology and editable assemblies. It targets BIM-style workflows where elements are named and carry explicit relationships, and it can hold the exact geometry of specified elements while leaving the rest flexible. Output exports to SketchUp (.SKP), Rhino (.3DM), and STL; AutoCAD and IFC are listed as planned. The intended audience and use cases are broad within design: product design, furniture, interiors, residential and urban architecture, landscape, complex structures, manufacturing, 3D printing, and game assets. It is a modeling and geometry tool — it produces files a designer can measure, cut, print, or machine, and hand off to the rest of a CAD pipeline. It does not write marketing copy, cut social video, design carousels, hold a brand voice, or schedule and publish anything. It builds the model; distribution and content are entirely outside its scope.
The clearest fit is architects, interior and product designers, and design studios who want to get from an idea, sketch, or reference to editable, precise geometry faster than starting cold in Rhino or SketchUp — especially those already living in a NURBS/CAD pipeline who need clean solids they can refine, not meshes they'd have to rebuild. Studios that iterate on early-stage concept models and value BIM-style named elements are the natural early adopters. It's a weaker fit today for anyone who needs production certainty: the pre-launch status (public preview opens September 18, 2026), absent pricing, planned-not-shipped AutoCAD/IFC support, and unverified output quality mean you should validate it against your own workflow before betting a deliverable on it. And it's the wrong tool entirely if your bottleneck is marketing the work — showing the design, winning the client, filling the pipeline — because Cartesian makes geometry, not content. For that side of a studio's job you'd pair it with a content engine like Kompozy.
| Dimension | Score | Why |
|---|---|---|
| Natural-language & multimodal input | 4.2 / 5 | Say-it / sketch-it / drag-in a reference is a genuinely low-friction on-ramp to a first model, and the multimodal capture (photos, plans, scans) is well-suited to design work. |
| Geometry precision (NURBS solids) | 4.3 / 5 | The headline claim — native NURBS solids and exact geometry rather than polygon-mesh blobs — is what designers actually need; unverified in preview, so confirm on your own output. |
| Editability & clean topology | 4.0 / 5 | Clean topology, editable assemblies, and the ability to lock exact geometry on specified elements while leaving the rest flexible is the right model for iterative design. |
| CAD / BIM interoperability | 3.7 / 5 | Exports to SketchUp (.SKP), Rhino (.3DM), and STL cover common pipelines, but AutoCAD and IFC are listed as planned rather than shipped. |
| Breadth of use cases | 4.0 / 5 | Spans product, furniture, interiors, architecture, landscape, manufacturing, 3D printing, and game assets — a wide net for a single modeler. |
| Maturity & availability | 2.8 / 5 | Hasn't launched yet — public preview opens September 18, 2026, behind early access and a waitlist; promising, but not a generally available product you can rely on yet. |
| Pricing transparency | 2.5 / 5 | No public pricing disclosed for Cartesian at review time, so cost and value are impossible to judge. |
| Usefulness for content / marketing | 1.5 / 5 | Not its job — it produces 3D files, not captions, social video, carousels, blogs, or scheduled posts, and it publishes nowhere. |
The most honest thing to say about Cartesian's pricing is that there isn't any to analyze yet. As of this review it hasn't launched — its public preview opens September 18, 2026 — with early access going first to Formas' founding-circle and studio members and a waitlist for everyone else, and no public price for Cartesian itself. Formas Studio, the company's separate AI rendering workspace, runs on a monthly credit model with a Pro tier — a reasonable hint at how Formas prices, but not a number you can plan around for Cartesian.
That makes value impossible to score fairly, so treat the "pricing" dimension as a placeholder that will move once Formas publishes rates. The relevant comparison for a designer is against the incumbents Cartesian exports to: Rhino is a one-time perpetual license, SketchUp is a subscription, and both are mature, proven tools. Cartesian's argument has to be that AI-driven, natural-language modeling saves enough time to justify whatever it eventually charges on top of — or instead of — those.
None of this is a knock; it's just early. But it does mean you should not commit a billable deliverable to Cartesian on the strength of a demo. Validate the geometry quality and export fidelity in your own pipeline first, and wait for published pricing before you model the cost side.
| Use case | Fit | Why |
|---|---|---|
| Early-stage concept modeling for architecture | Strong | Turning a sketch or reference into an editable model fast is exactly the friction Cartesian targets, and the NURBS-solid claim fits arch pipelines. |
| Product and furniture design with precise geometry | Strong | Exact solids and editable assemblies are what product designers need, and STL export feeds 3D printing directly. |
| Studios already living in Rhino or SketchUp | OK | Native .3DM and .SKP exports mean output drops into existing pipelines, though the tool is still a preview to validate. |
| BIM workflows needing named elements and documentation | OK | Named elements with explicit relationships point at BIM, but IFC and AutoCAD support are planned rather than shipped. |
| Photorealistic client renders | Weak | Cartesian builds geometry; you still need a renderer (Formas Studio or another tool) to produce presentation visuals. |
| Marketing the design work across social platforms | Weak | It makes no video, carousels, captions, or posts and publishes nowhere — that is a content engine's job, not a modeler's. |
| Production deliverables under a deadline | Weak | Pre-launch status (preview opens September 18, 2026), no pricing, and unverified output make it risky to bet a due deliverable on right now. |
Scored on its own terms Cartesian is a credible early AI modeler, and Kompozy has no business pretending to compete with it — one builds 3D geometry, the other runs a content operation. The place they meet is the part of a design studio's job Cartesian doesn't touch: turning finished work into visible marketing that wins the next client. A precise NURBS solid sitting in a .3DM file makes nobody a lead. A studio still has to show the design, tell the story of how it was made, and put it in front of the people who commission work — and that's a content problem, not a modeling one.
That's where Kompozy fits. Screen-record a turntable of a Cartesian model, or export a render, and bring it into Kompozy: it becomes vertical [Clipped Shorts](/glossary/clipped-short) with branded captions and per-platform reframes, a brand-exact process carousel via [HyperFrames](/glossary/hyperframes) walking through sketch-to-model, quote graphics, a project case-study blog article, and an email newsletter to your list — all held to your studio's voice by a [Persona Brief](/glossary/persona-brief) and scheduled across the eight social platforms plus blog and email with [Autopilot](/glossary/autopilot). You can even put a recurring, face-locked presenter in front of the reveal with [Persona Shorts](/glossary/persona-shorts). The honest framing isn't "which is better" — it's "model it in Cartesian, market it with Kompozy." They sit on opposite ends of the same studio workflow.
As an AI 3D modeler for architects and product designers, it's a promising early tool — it aims at precise, editable NURBS solids and exports to Rhino, SketchUp, and STL, which is what a CAD pipeline actually needs. But it hasn't launched yet (public preview opens September 18, 2026), has no published pricing, and its output claims are unverified, so it's worth trying and validating on your own work once the preview opens rather than committing a billable deliverable to it now.
Cartesian is an AI 3D-modeling tool from Formas (formas.ai) that turns natural-language descriptions, sketches, and dragged-in references into editable 3D models. It targets architecture and product design, claims native NURBS solids with clean topology, and exports to SketchUp (.SKP), Rhino (.3DM), and STL, with AutoCAD and IFC listed as planned.
Most text-to-3D tools produce polygon meshes aimed at game assets or quick visualizations. Cartesian claims native NURBS solids and precise, editable geometry aimed at CAD and BIM workflows — the difference between a mesh you'd have to rebuild and a solid you can measure, cut, and machine. That precision claim is unverified in preview, so confirm it on your own output.
Formas hasn't published pricing for Cartesian at review time — its public preview opens September 18, 2026, with early access going first to founding-circle and studio members and a waitlist for others. Formas Studio, the company's separate AI rendering workspace, uses a monthly credit model, but that's not a stated price for Cartesian. Check the Formas site for current terms.
No. Cartesian builds 3D models and geometry — it writes no captions, cuts no social video, designs no carousels, and publishes nowhere. To turn a model or render into marketing across platforms, you'd use a content engine like Kompozy, which clips, captions, repurposes, and publishes.
Cartesian exports to SketchUp (.SKP), Rhino (.3DM), and STL, so its output drops into common design pipelines and 3D-printing workflows. AutoCAD and IFC support are listed as planned rather than currently shipped.
No. Formas Studio is Formas' AI rendering and visualization workspace (image generation, 3D-from-image, architectural visualization) on a credit model. Cartesian is a separate, CAD-adjacent product focused on generating precise, editable 3D models from language, sketches, and references.
They do different jobs. Use Cartesian to build the 3D model; use Kompozy to market it — turning a turntable capture or render into clipped shorts, a process carousel, a case-study blog, and a newsletter in your brand voice, then publishing across nine platforms. Many studios model in Cartesian and run their published brand in Kompozy.
See Cartesian by Formas vs Kompozy comparison → · Get Started →