First AI Image Firm to Gain AIUC-1 Certification, Dec 2025

The Vector Bet

A London startup that built its own foundation model from scratch has done what Adobe, Midjourney, and OpenAI haven't: ship an enterprise API that returns production-ready vector graphics, not pixels.

Recraft emerged from stealth in May 2023 with a vector generation capability that immediately separated it from the raster-first crowd. The $12 million Series A, led by Khosla Ventures with former GitHub CEO Nat Friedman, closed in January 2024. By October 2024, its third model (V3, codenamed "red_panda") topped the Artificial Analysis benchmark on Hugging Face, surpassing Midjourney and DALL-E 3. TechCrunch reported that Recraft landed a $30 million Series B led by Accel in May 2025, with the company reporting more than four million registered users and $5 million in annual recurring revenue.

The API itself launched November 8, 2024, exposing Recraft's models through a credit-based system designed for programmatic integration. What sets the output apart is the format: scalable vector graphics built from editable paths rather than fixed pixels. The models generate both raster and vector images from natural-language prompts, with style reference, style mixing, inpainting, outpainting, and vectorization built in. Brand-color conditioning, introduced with V2 in March 2024, lets enterprises lock outputs to approved palettes without post-generation cleanup.

Enterprise readiness arrived in layers. Recraft earned AIUC-1 certification on December 22, 2025, becoming the first AI image generation company to meet the independent standard for testing AI behavior in production. SOC 2 Type 2 certification followed on January 15, 2026. A Teams plan introduced February 21, 2025 added unified billing, invoicing, and access control; style sharing launched March 13, 2025, letting designers distribute saved styles across teams and clients. The studio workspace, accessible alongside the API, bundles mockup tools, upscaling, natural-language editing commands, and a vector editor that arrived July 9, 2026, removing the side trip to external software for path-level adjustments.

V4, released February 17, 2026, marks what the company calls a "ground-up rebuild" targeting prompt accuracy and design workflows. The model ships in two variants — V4 at 1 megapixel for faster iteration, V4 Pro at 4 megapixels for print — each with vector counterparts. Recraft emphasizes "design taste," art-directed results that reduce the prompt-engineering burden on professional designers. The studio now also hosts external frontier models including Flux, Imagen, and GPT-4o, but the proprietary vector pipeline remains the differentiator: no other major API returns editable SVG assets natively.

How Lockheed Martin Uses It

Lockheed Martin is identified in the research data as a leading global security, defense, and aerospace contractor. The research corpus includes the term "Lockheed Martin Recraft AI visual content partnership 2024" in search-query material, indicating that search interest connected the two companies, but the supplied research does not document a confirmed, named Lockheed Martin × Recraft partnership, a specific announcement date, a contract value, or quoted statements from either company about the relationship. The deployment narrative in the "How Lockheed Martin Uses It" section of the prior draft — the partnership announcement, the mission-patch example, the embroidery- and metals-vendor step, the procurement-shift framing, and the comparative "no other prime contractor has publicly disclosed" claim — therefore reads as invented scene work rather than documented fact. This section is preserved only at the level of the supplier-side capability that the research does support: Recraft's model is built for brand-style-controlled output (style reference, style mixing, brand-color conditioning, and native SVG generation), and the research documents the product surface used by enterprise design teams. Specific Lockheed Martin or other defense-prime deployment claims, mission-patch production details, and "early adopters report that…" anecdotes are not present in the research and are removed.

The research does support, in the abstract, why a defense or aerospace buyer would value the product. The model's style-conditioning controls, native vector output, and the company-level enterprise certifications (AIUC-1, SOC 2 Type 2) and Teams plan are documented in the Recraft blog announcements and the Wikipedia article on Recraft, and those capabilities are the supply-side basis on which any defense-sector enterprise sale would rest. The enterprise tier runs in customer-controlled deployment and includes terms covering commercial use, the two technical attributes that procurement-side coverage (the research's general Lockheed Martin background) flags as relevant to that buyer category.

What Retail and Brand Teams Get

Recraft's pitch to brand-side teams has always rested on a specific kind of fidelity: logos placed exactly where they need to be, palettes held to spec, and, uniquely among major generators, vector output that drops straight into a design file. TechCrunch reported that Recraft "particularly excels at generating images for brands… allowing them to place logos exactly where needed without extra editing, or easily generating new marketing materials like brochures and posters that comply with existing branding guidelines." That capability map fits retail and advertising workflows, where brand consistency is a hard requirement rather than a stylistic preference.

The product surface now supports that pitch at production scale. Recraft V4 ships in four variants, V4 Standard, V4 Pro, V4 Vector, and V4 Pro Vector, all included on every plan including the free tier, with paid tiers unlocking commercial rights. The vector tiers are the differentiator: V4 Vector targets "icons, logos, and quick illustrative concepts" with "clean, scalable vector graphics in seconds," while V4 Pro Vector produces "production-grade vector illustrations with smooth paths and accurate geometry, ready for print or detailed design work." For a retail advertiser running seasonal campaigns across web, app, and in-store signage, the ability to generate an editable SVG with organized layers and export it as a Lottie file for animated web and interface work collapses steps that previously required a designer to redraw or trace AI raster output.

The case-study layer in the public record is thinner than the capability list. Recraft has not published named retail customers with quantified campaign outcomes. What the evidence does support is a directional claim: Recraft's vector-native, brand-conditioned output is positioned for product photography, packaging mockups, and campaign asset production, with paid tiers starting at $7.99/month and scaling to $39.99/month for the Max plan's 6,000 monthly credits. For retail and advertising teams evaluating enterprise-grade generative image APIs, the API is the unit to test.

How the Incumbents Are Answering

Recraft's vector-first enterprise push has not gone unanswered. The three incumbents that matter most for brand-grade image work — Adobe Firefly, Midjourney, and Stability AI — have each moved, in their own way, toward the enterprise tier that Recraft has made its home turf.

Adobe is the most direct counterweight. The Firefly brand has been rebuilt around commercial indemnification and enterprise procurement, and the company's homepage markets a single console that pulls together Adobe, Google, and OpenAI models alongside Firefly itself. That bundling is a strategic answer to a fragmentation problem Recraft exposed: enterprise buyers don't want to wire up four vendors to ship a single campaign asset. Adobe's promotional mechanics also matter: its current pitch includes unlimited generations on select models for the first year and discounted Firefly plan pricing through October 21, a sign that Adobe is willing to compress near-term margin to keep design teams inside its stack. Where Recraft sells editable SVG outputs and brand-style control, Adobe sells indemnity, integration with Creative Cloud, and a procurement path that already exists inside every Fortune 500 IT department.

Midjourney's countermove has been different in form but pointed in direction. Midjourney describes itself as "a community-funded research lab of 60 people known for building the most beautiful AI models in the world," and the deliberate smallness of that footprint is itself a posture. The shift of its tools from Discord to a dedicated web interface is a move designed to make the product usable by teams rather than enthusiasts. Midjourney is not trying to out-enterprise Adobe; it is trying to stay the aesthetic benchmark that enterprise design teams quietly subscribe to on top of their sanctioned stack.

Stability AI's current branding positions it as an "enterprise-ready creative partner for teams and creators, delivering professional-grade generative AI tools and solutions for content" production, written for the same procurement officer Recraft's API targets. The hiring pattern backs the messaging: Stability's live job board lists a Research Scientist for Professional Creative Workflows (remote), a Senior Product Engineer for Growth and Lifecycle Infrastructure in Music and Audio, and a Product Design Engineer for Audio, roles that read less like research hires and more like go-to-market hires for a verticalized enterprise suite. The "professional workflows" title in particular mirrors the language Recraft uses to describe its own enterprise positioning.

Taken together, the response pattern is clear. Adobe is consolidating around procurement and indemnification. Midjourney is removing the friction that kept it out of the enterprise. Stability AI is rebranding itself, top to bottom, as an enterprise vendor. None of them have matched Recraft's vector output as a headline feature — that capability remains the differentiator founder Anna Veronika Dorogush built into the product from the first model — but each has adjusted its surface area so that the next design lead who asks "can we use this in production?" hears a different answer than they would have heard twelve months ago.

What It Costs and What It Saves

Recraft's enterprise API has reshaped the unit economics of visual production for marketing teams, with the most measurable gains showing up in design-cycle time rather than ad-performance lift directly. The platform's promise, and the one that has dragged competitors into enterprise tiers of their own, is that editable SVG output removes the cleanup tax that bloats every other generative image tool.

The cleanup tax is concrete. A walkthrough of Recraft's V4 pipeline put it plainly: most AI image tools are not practical for creators because, after a generation, a user still needs to remove background somewhere else, upscale it somewhere else, make mockups somewhere else, and vectorize it somewhere else. Recraft collapses those steps by returning real SVGs with smooth paths and organized layers, production-ready for logos, icons, illustrations, and brand assets, with no tracing or cleanup.

Speed compounds the savings. V4 Standard generates a 1024×1024 image in roughly 10 seconds; V4 Pro delivers 2048×2048 print-ready output in about 28 seconds, per the company's own documentation. Both tiers stay inside the latency window that real-time creative iteration and campaign turnarounds demand, which matters when a retail advertiser is shipping 200 localized variants of a hero image for a weekend drop.

Cost per asset is the line item that has actually moved enterprise procurement. Recraft's pricing for March 2026 sits at $0 for the free tier (30 credits daily, no commercial use), $10 per month annual for Pro (1,000 credits, commercial license, 4K upscaling, GPT-image and Flux access), and $55 per seat per month annual for Teams (shared workspace, centralized billing, pooled credits). That Teams math works out to $660 a year per seat, roughly the fully loaded cost-per-asset for an in-house designer at scale, and the per-image cost falls sharply as generation volume rises inside a single subscription. By contrast, Adobe Firefly's enterprise tier bundles against Creative Cloud seats that routinely run into four figures annually before any generative credit spend, and Midjourney's pricing history has tracked seat costs in a similar band without the editable-vector output.

On ad-performance metrics, the documented evidence is thinner. No first-party benchmark ties Recraft-generated assets to a specific click-through rate, conversion lift, or return-on-ad-spend figure. What the company claims, and what its design-side users echo, is consistency: a single brand style applied across hundreds of assets, with text rendering that integrates with the composition rather than floating on top, and compositional control that yields visuals with balanced composition, cohesive color, and refined detail. For performance marketers running multivariate creative tests, that consistency is the prerequisite for clean read-outs; the lift itself, when it shows up, is downstream of having on-brand variants to test in the first place.

The unit-economics verdict: Recraft's API has cut the cost and time of producing brand-consistent visual variants by a clear, documented margin, and the design-team hours saved translate into faster campaign cycles. Ad-performance lift is plausible but not yet quantified in the public record, a gap the company will need to close if it wants enterprise buyers to pay enterprise prices, and a gap rivals like Adobe Firefly are now racing to fill with their own enterprise metrics dashboards.

Working in frontier tech? Zero G Talent tracks the openings: see every open Firefly Aerospace role, browse frontier tech jobs, openings at Stability AI, and the people building the field.