Claude Fable 5
via claude-code-cli 2.1.214model + harness
Run of 18 Jul 2026
2 format checks passed — no check of the result
2 render and interaction checks not run
Evaluation not published
Archived pilot · Historical run
Aurora/Mesh-Gradient-Art
Visual craftbuilt by Phi-4-mini-instruct via ollama 0.32.15 model only
- Automated checks
- 1 of 2 format checks passed — no check of the result2 without a recorded result
- Judge
- Evaluation not published
- Started
- Duration
- 12 s
Artifact
Rendered by agi.jetzt from the model-produced artifactRendered offline in an isolated renderer from the stored file (sha256 5ef7c9757617…). The file itself is not served.
Same brief, other models
Scroll sideways. Each run had its own runtime and limits — compare the artifacts, not a score.
GPT-5.6 Luna
via codex-cli 0.144.5model + harness
Run of 18 Jul 2026
2 format checks passed — no check of the result
2 render and interaction checks not run
Evaluation not published
Phi-4-mini-instruct
via ollama 0.32.15model only
Run of 30 Aug 2026token limit 16000
1 of 2 format checks passed — no check of the result
2 render and interaction checks not run
Evaluation not published
Qwen3.8-27B
via ollama 0.32.15model only
Run of 3 Sept 2026
2 format checks passed — no check of the result
2 render and interaction checks not run
Evaluation not published
Automated checks
Automated checks of the stored HTML file, without running it in a browser. Details are the runner’s German originals. Format and presence checks say nothing about whether the result fulfils the brief.
| Check | Kind | Result |
|---|---|---|
| single_html5_filedetail not recorded | Format | Failed |
| no_external_requestsdetail not recorded | Rendering | Unknown |
| renders_without_console_errorsdetail not recorded | Rendering | Unknown |
| viewport_metadetail not recorded | Format | Passed |
2 render and interaction checks are recorded without a result; whether they ran is not recorded.
Evaluation
- Deterministic score
- No score — only the format was checked, not the result
- Judge
- not published for this run
WhyNo judge evaluation was published for this run.
Limits of this run
- No executed check with a shown result evaluates what was built.
- No judge evaluation is published for this run. There is therefore no quality verdict — only the automated checks listed here.
- 2 checks have no recorded result. Unknown stays unknown — it is not counted as failed.
- Token limit 8192: other runs of this task were recorded with 16000, so this run is not like-for-like with them.
- Not fully reproducible; the recorded reason is listed under “Environment”.
Task
Brief (turn 1)
Baue als EINE einzelne HTML-Datei (kein Framework, kein externes Asset, keine externen Fonts) einen bildschirmfüllenden, animierten Hintergrund im Stil Biolumineszenz in der Tiefe: tiefes Petrol bis Schwarz, darüber schwebende, leuchtende Schlieren in Türkis, Cyan und einem kalten Magenta, die sich träge wie Quallen bewegen. Geheimnisvoll, ruhig, kühl. Helle, klare Wortmarke. für die Startseite eines Tiefsee-Forschungsprojekts.
Anforderungen:
- Ein atmosphärischer, animierter Verlauf (Aurora / Mesh-Gradient / fließende Farbwolken), der den gesamten Viewport füllt und sich langsam, organisch bewegt — via CSS-Animation oder Canvas, selbst erzeugt.
- Ruhige, hochwertige Bewegung mit weichem Easing; die Farben durchmischen sich stimmig, kein hartes Flackern, keine sichtbaren Kanten oder Banding-Streifen.
- Darüber eine knappe, zentrierte Bildmarke: „Abyssal" als elegante Wortmarke plus eine kurze, ruhige Zeile darunter — gut lesbar über dem bewegten Grund.
- `@media (prefers-reduced-motion: reduce)`: Bewegung anhalten, ein ebenso schöner statischer Verlauf bleibt.
- Bildschirmfüllend ohne Scrollbalken, kein horizontaler Overflow.From the task seed (German original); the actual prompts add output-format instructions and other wording.
Environment
- Runtime
- ollama 0.32.15
- Execution
- model onlyThe runner reached the model directly — no agent harness, no tools.
- Profile
- v3-local-phi4-8192-v1 · T=0.2 · seed 7331 · 1 turn · no tools
- Max tokens
- 8192
- Reproducibility
- not fully reproducible8192-token execution profile differs from historical 16000-token profile; upstream weight revision remains unknown
- Run window (runner clock)
- 2026-10-01T18:35:10.313+00:00 → 2026-10-01T18:35:21.98+00:00
Provenance
- Source system
- agi.jetzt v3 (Prüfstand worker)
- Source run
- 55b6c305-9d82-4baa-917d-43f1a960e2f2
- Status
- submitted by the Prüfstand worker (v3)
- Run manifest (sha256)
- not recorded
- Artifact (sha256)
- 5ef7c9757617be3adda0f99b662d2ba711dbe73c37a55b7286dfeec8ebe31d4c
- Checks version
- not recorded
Model The runner’s model identifier maps to Phi-4-mini-instruct in the agi.jetzt register (verified mapping).
What this is evidence for
Not an AGI measureThis run shows one thing: what a model built as a single HTML file from a precise brief. It says nothing about general intelligence.Checked automatically in this run: 1 of 2 format checks passed — no check of the result · 2 without a recorded result.
What the task set is meant to measure
- Artifact creation from a precise brief
- Instruction following
- Static checks of the file (no functional test)
Not in its scope
- General intelligence / AGI
- Agents and tool use
- Multimodal input
- Retrieval
- Conversation
Published and unpublished runs
4 of 15 verified v2 runs of this task are published. 4 of the 6 models with recorded runs have a published run.Not published: Claude Haiku 4.5 (3 runs; 0 of 2 format checks passed); GPT-5.6 Sol (3 runs; 0 of 2 format checks passed).In addition, 1 published run has no selection record (not part of the v2 import).Source: the selection records of the published runs. They count the task across its variants and know only models with at least one published run.
This run has no selection record (it is not part of the v2 import). How many further runs of this model exist is not recorded.