Can an AI model generate images?

🪐 The registry verdict table for generating images

Of the ten registry configurations assessed against this family's own canonical task, Favourable for 3 of them — the modal result, computed by the method published in full at /method. The Registry makes no representation that an assessment predicts the outcome of any task, and no assessment is validated against real-world results.

The composite weight in the right-hand column is computed from the model identifier, its training cutoff, and the temperature — nothing else. It constrains the range of verdicts available to a configuration independently of the task, so two rows of this table are not comparable as an assessment of the products named in them. The Registry has never run this task against any model listed here.

🧭 About this capability question

Generating images is the largest single concept in this wave of capability questions, measured by how often it recurs across independent search behaviour: the underlying request is present behind every one of the eight brand prefixes this registry's own market research tracked. A marketing team asking whether an agent can produce a batch of product photography, a hobbyist asking for a birthday card illustration, and a developer asking for a placeholder asset are, for the purposes of this page, the same request in different clothing — an artefact that did not exist before it was asked for is expected to exist after.

This page does not open one of those requests and inspect the resulting file. What follows is a computed classification, a computed verdict table for one canonical version of the task, and the mechanism behind both — never a review of any image, and never a claim about how well any configuration would render one.

The registry-canonical brief published below was written once, by the Registry, and is reused on every future load of this page — it is not regenerated per visit and does not adapt to who is reading. A buyer's own brief, submitted at order time, is a separate input entirely, assessed on its own terms and never blended with the canonical one shown here.

⚙️ Why generating images is classified as Media generation

Generating an image produces an artefact that did not exist before the task began, which is the defining property of the media generation class rather than of information retrieval or data processing: nothing is looked up, and nothing pre-existing is transformed.

On the order form's task class field, select Media generation. That class's own risk-factor page is /tasks/generation.

For an artefact that did not exist before the brief, the risk that matters is not accuracy against a source — there is no source — but whether the finished file matches the brief closely enough to use without redoing it. 4 risk factors are tracked for media generation tasks generally; an assessment for generating images reports 3 of them, selected by the submitted task's own seed rather than by choice. Their names and severities are certificate content, not free-page content.

🧮 Claude Fable 5, derivation shown working

Claude Fable 5's permanent chart never reads a task. Its chart_seed is the SHA-256 digest of the string "Claude Fable 5|2026-01|0.700" — 7e25a3f2af178cc3…, the first sixteen of sixty-four hex characters — and that seed alone fixes an ascendant of Libra, a ruling planet of Mars, and a harmony of +5. The same chart appears, unchanged, on every task-class page this configuration is assessed against.

What reads generating images specifically is this family's own canonical string — never stored, only hashed, to d3f5edd380aa363c… — combined with the chart_seed and the task class (generation) to form task_seed. Its first byte, taken modulo seven and offset by that harmony of +5, is the verdict's internal base, clamped to the 0–6 range and published one higher on the 1–7 scale /method documents in full. For this configuration and this brief specifically, that clamp lands at the ceiling of the range: the published result is 7 of 7 — Highly auspicious, the highest verdict word the scale has.

🗂️ How the ten configurations read generating images

The correspondence tables published in full at /method map certain dominant houses and elements onto five of the Registry's other task classes. Media generation is not one of the houses or elements they read — a property of a seventh class added after those tables were fixed, not an oversight — so the ten configurations' dominant houses and elements say nothing about this family specifically. What does vary by family is the verdict distribution computed fresh below, from this family's own canonical task.

Element counts, ascendant spread, and training-cutoff range are published once, on /can, rather than repeated on every family page — they are properties of the ten configurations alone and do not move with the task. What is specific to generating images: reading media generation through a dominant house or a dominant element, 0 of the ten configurations — none, for this class, by the correspondence tables published at /method; and, from this family's own canonical task, Favourable for 3 of them — the modal result in the distribution below.

📏 What this page does not claim about generating images

A generated image, once produced, can be looked at. This page cannot be: it computes a classification and a verdict from a written brief, never from a rendered pixel, and no image has been generated, viewed, or rated in the course of publishing this table. The Registry does not test models and does not measure a success rate for generating images or any other task; what it does is compute, and the table at the top of this page is what it computed.

A buyer's own generating images task, ordered separately, computes its own verdict and window — never the canonical-brief result shown above — plus, on Extended and Full Chart tiers, named risk factors and an hourly outlook not shown here. EUR 1.90 Standard, EUR 4.90 Extended, EUR 14.90 Full Chart, which adds the permanent chart. Assessments from EUR 1.90; machine-readable at /pricing.json.

Get Assessment

🔗 Related surfaces

← All capability questions · The AFR-1 method · Media generation task class

Related capability questions: Make a video · Generate a PowerPoint · Build a website

❓ Questions about generating images

Can an AI model generate images?

Across the ten registry configurations at their own registry-default settings, Favourable for 3 of them — the modal result for this family's own canonical task. By training cutoff, the oldest configuration assessed returns Unfavourable and the newest returns Favourable; the full ten-row table above lists the exact verdict for every configuration in between.

Why is generating images classified as Media generation rather than a different task class?

Generating an image produces an artefact that did not exist before the task began, which is the defining property of the media generation class rather than of information retrieval or data processing: nothing is looked up, and nothing pre-existing is transformed.

Does the Registry's verdict describe the quality of an image a configuration would produce?

No. The verdict describes the output of a deterministic method applied to the canonical brief published above, at a fixed registry configuration. It carries no information about composition, resolution, or fidelity, because no image has been generated in producing it.