Open Weight Bench

nvidia/nemotron-3-nano

nvidia 30B · 3B active mlx / 4bit ctx 256k released 2025-12-15 tool_use
Score
35%
Static
100%
Functional
33%
Qualitative
31%
tok/s
131
Tokens
5504
Wall
69.1 s
RAM
16.6 GB
Worum geht es? Was wird getestet?
Task: From a ~200-word prompt the model must generate a fully functional Kanban board as a single-file HTML with drag & drop, localStorage persistence, edit/delete and a confetti animation — in a single chat without iteration. The prompt also includes a small `data-testid` contract so a Playwright test can drive the app remotely. Three signals feed into the score: (1) Static — a linter checks concrete constraints in the HTML (columns, Tailwind, localStorage call, no framework, no window.alert/prompt, …). (2) Functional — Playwright runs a small CRUD sequence: create a card, delete a card with confirmation, reload — does state persist? — and checks whether any JS console errors occur during the entire flow. Drag & drop and confetti are deliberately not tested functionally (too many implementation variants). (3) Qualitative — LLM-as-judge rates screenshot and code (visual + code quality + render↔code consistency). Score = mean over the available signals. Why models fail: reasoning models burn their tokens in thinking instead of writing. Sliding-window models (Gemma 4) lose the constraints at the start of the prompt. Small models (<3B) often fail to produce coherent HTML — or ignore the data-testid contract, which makes the functional tests fail in droves.
Prompt
System prompt
You are a careful front-end engineer.
Developer prompt
Create a fully functional Kanban board in a single HTML file using vanilla JavaScript (no frameworks like react).

Requirements:
- Columns: Backlog, In Progress, Review, Done.
- Cards must be:
  - draggable across columns,
  - editable in place,
  - persisted in localStorage (state survives reloads) - please use your own namespace,
  - deletable with a confirmation prompt.
- Each column provides an "Add card" action.
- Style with Tailwind via CDN.
- Add subtle CSS transitions and trigger a confetti animation when a card moves to "Done".
- Thoroughly comment the code.
- dont use window.alert or window.prompt to add/edit/delete cards
- if there are no cards yet, create some dummy cards
- modern and vibrant design

Stable test selectors (mandatory — these data-testid attributes are used by an automated functional test; do not omit, rename, or split them across multiple elements):
- Column containers: data-testid="column-backlog", data-testid="column-in-progress", data-testid="column-review", data-testid="column-done".
- Every "Add card" button (one per column): data-testid="add-card".
- Every card element: data-testid="card".
- Inside each card, the delete trigger: data-testid="delete-card".
- The confirm button of the delete-confirmation dialog/modal: data-testid="confirm-delete".
- The input/textarea where a new card title is typed: data-testid="card-input". Pressing Enter in this input MUST commit the new card.

As answer return the plain HTML of the working application (script and styles included)
▶ Open generated app in a new tab ↗
Screenshot der gerenderten App
Screenshot der gerenderten App
Qualitative · LLM-as-judge (openai/gpt-5.4) 2026-04-29T20:59:38.407421+00:00
31%
Visual (screenshot)
  • board renders
    50%
  • column completeness
    100%
  • cards present
    0%
  • ui affordances
    50%
  • design quality
    40%
Die vier Spalten sind klar sichtbar und die Add-Card-Buttons sowie Eingabefelder werden gerendert. Es sind jedoch keine Karten sichtbar, stattdessen ein offen stehender Delete-Dialog, wodurch das Board unfertig und funktional gestört wirkt.
Code quality (HTML/JS)
  • code structure
    50%
  • dom safety
    20%
  • robustness
    10%
  • code quality
    20%
Der Code ist grundsätzlich in Funktionsblöcke gegliedert, enthält aber mehrere gravierende Defekte: inkonsistente Schlüssel zwischen DOM-Dataset und State, falscher Modal-Selector für den Confirm-Button und unsicheres innerHTML für Kartentitel. Robustheit ist schwach, da localStorage/JSON.parse ohne Absicherung genutzt werden und mehrere Stellen zur Laufzeit brechen können.
Render ↔ code consistency 0%
Starke Diskrepanz zwischen Anspruch und Render: Der Code seeded Dummy-Karten und hat Renderlogik, im Screenshot erscheinen trotzdem keine Karten. Das passt zu den offensichtlichen Laufzeit-/Mapping-Fehlern im Code, die das Initial-Rendering oder die Interaktion faktisch brechen.
Static · linter checks (12/12 passed)
  • Spalte 'Backlog'
  • Spalte 'In Progress'
  • Spalte 'Review'
  • Spalte 'Done'
  • HTML5 drag & drop or sortable lib
  • localStorage persistence
  • Tailwind via CDN
  • Confetti animation on 'Done'
  • Add card button
  • Vanilla JS (no React/Vue/Angular/Svelte)
  • No window.alert/prompt for CRUD
  • Editable cards (contenteditable / input)
Functional · E2E tests (2/6 passed)
  • Page loads with all 4 columns
  • Add card via button
    card was not created (count unchanged)
  • Card is added exactly 1× (no double submit)
    no add happened, so not verifiable
  • Delete card via button + confirmation
    no cards available to delete
  • Cards survive a reload
  • No JS console errors
    pageerror: Cannot read properties of null (reading 'addEventListener'); pageerror: Cannot read properties of null (reading 'addEventListener')
Raw model response
```html
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<title>Kanban Board</title>
<script src="https://cdn.tailwindcss.com"></script>
<style>
  body { background: #f0f4f8; font-family: system-ui, sans-serif; }
  .column { @apply p-4 border rounded-md;
    transition: transform 0.2s ease, box-shadow 0.2s ease;
  }
  .column:hover { transform: translateY(-2px); box-shadow: 0 4px 12px rgba(0,0,0,0.1); }
  .card { @apply flex items-center justify-between p-3 bg-white rounded shadow-sm hover:shadow-md transition; }
  .card-title { @apply flex-1 mr-2 cursor-pointer double-click:underline; }
  .card-content { @apply flex items-center; }
  .delete-btn { @apply text-red-600 hover:text-red-800; }
  .input-new { @apply w-full p-2 border rounded-md mb-2; }
  .modal { @apply fixed inset-0 bg-black bg-opacity-50 flex items-center justify-center z-50 hidden; }
  .modal-content { @apply bg-white rounded p-4 w-64 md:w-56; }
  .confetti { @apply fixed w-2 h-2 rounded animate-fall; }
  @keyframes fall {
    0% { transform: translateY(0) rotate(0deg); }
    100% { transform: translateY(-30vh) rotate(360deg); }
  }
</style>
</head>
<body class="min-h-screen flex flex-col items-center p-4">

<!-- Columns Container -->
<div class="grid grid-cols-4 gap-4">

  <!-- Backlog -->
  <div class="column" data-testid="column-backlog">
    <h2 class="text-xl font-semibold mb-2">Backlog</h2>
    <button data-testid="add-card" class="w-full bg-indigo-600 text-white py-1 rounded mb-2">Add Card</button>
    <input type="text" class="input-new" data-testid="card-input" placeholder="New card title">
    <div class="cards space-y-2"></div>
  </div>

  <!-- In Progress -->
  <div class="column" data-testid="column-in-progress">
    <h2 class="text-xl font-semibold mb-2">In Progress</h2>
    <button data-testid="add-card" class="w-full bg-green-600 text-white py-1 rounded mb-2">Add Card</button>
    <input type="text" class="input-new" data-testid="card-input" placeholder="New card title">
Score
69%
tok/s
86
Tokens
11228
Wall
379.0 s
RAM
16.6 GB
Worum geht es? Was wird getestet?
Task: In a German book corpus (with embedded source code) 10 synthetic facts are hidden at evenly distributed depths (5% – 95%). The model must retrieve all of them. Flow — THREE turns in the same chat context (prefill only once): Turn 1 — corpus summary: model receives the long corpus and summarises it in 3-5 sentences. Forces real processing of the text. Turn 2 — needle retrieval: same conversation, now the questions for the 10 hidden facts. Turn 3 — comprehension + hallucination traps: 6 questions about the book (4 factual + 2 traps where the answer is NOT in the text — the model should recognise this rather than fabricate). Default mode runs ONE uniform stage for all models: 120k tokens. Models without sufficient max_context are skipped at this stage. `niah_deep` additionally runs 32k / 64k / 200k for a full heatmap. Score weighting: summary 20% + needle retrieval 50% + comprehension/hallucination resistance 30%. Why models fail: sliding-window attention (Gemma 4) only sees the last 1-2k tokens sharply. Reasoning models hit the token limit before answering. Q4 KV cache measurably degrades recall at long contexts. On the hallucination traps the helpful bias lures models into plausible-sounding inventions.
Prompt
Developer prompt
TURN 1 (User):
The following section contains a longer mixed text of German narrative and source code.

===== TEXT BEGIN =====
<corpus with embedded needles, 32k–128k tokens depending on stage>
===== TEXT END =====

Summarise the text in 3-5 sentences. Mention the main characters, setting and key themes.

TURN 2 (User, same chat context):
Now answer the following questions strictly from the text shown above — invent nothing, add nothing and do not rely on general knowledge.

Questions:
1. <question for needle 1>
2. <question for needle 2>
...

Answer as a numbered list 1., 2., 3. with one short sentence each.

For each context length 10 needles are distributed across the corpus. The NIAH score combines corpus summary, needle retrieval and optionally comprehension + hallucination traps. LLM judges replace the heuristic scores for summary and comprehension whenever available.

Stage: 120k tokens
Sub-benchmark · corpus summary Turn 1
45% LLM judge heur. 100% ~120k tokens input
**Zusammenfassung (3‑5 Sätze)**  

Der Text schildert das Leben des jungen Schmiedelehrlings Gottlieb Lasso in einem idyllischen, fast märchenhaften Blumental im preußischen Provinz­land, das von den Napoleon‑Kriegen und den rauen Alltag der Bauern, Schmiede und Fischer durchzogen ist. Hauptfiguren sind Gottlieb Lasso, seine Freundin Malineken (Tochter des Fischers Werpe), die alte Großmutter, der Schmied Michael Lebbin und die geheimnisvolle Prinzessin vom See (später als Gräfin Barnewitz/Comtesse Barnewitz enthüllt). Schauplatz ist das Blumental‑Wald‑ und Seegelände um Wriezen, das Dorf Finkenwalde und die umliegenden Höfe, während die Handlung zwischen Werkstatt, Wald, Schiffs­kahn und einer verborgenen Waffen­keller‑Krypta wechselt. Kernthemen sind der Konflikt zwischen deutscher Heimatverbundenheit und französischer Besatzung, das Ringen um Ehre, Rache und Freiheit, die Rolle von Glaube und Pflicht, sowie die symbolische Suche nach einem „Schlüssel“ – sowohl zu Waffen als auch zu einem möglichen Befreiungs­pfad für Gottlieb.
LLM judge summary (openai/gpt-5.4) 45%
  • main characters
    60%
  • setting
    85%
  • plot
    20%
  • themes
    80%
  • no hallucinations
    20%
Schauplatz, Zeitkontext und einige Figuren stimmen teilweise, ebenso die Themen Rache, Glaube und Besatzung. Die Zusammenfassung halluziniert aber mehrere zentrale Elemente: eine Großmutter, die Comtesse/Gräfin Barnewitz, einen Fischer Werpe, eine Waffenkeller-Krypta und einen „Schlüssel“-Plot; außerdem verfehlt sie den eigentlichen Handlungsbogen um Kriegskassen, Verrat und christliche Vergebungsproblematik.
  • ✓ 3-5 sentences · 4 sentences counted
  • ✓ At most 220 words · 144 words
  • ✓ Corpus keywords (3/4) · ✓ Gottlieb, ✓ Malineken, ✓ Schmied, ✗ Bonaparte
Sub-benchmark · needle retrieval Turn 2
9/10 hits
  • 5% (5 %)
    expected: blauer Ankerstein, Lübeck-1907, A-318
  • 10% (10 %)
    expected: smaragdgrün, Schlüssel, Seriennummer, 7-Bravo-12
  • 15% (15 %)
    expected: Nordstern-Brigade, violetter
  • 25% (25 %)
    expected: Indigo-Quark
  • 33% (33 %)
    expected: Atlantis-Mira, NL-7711, 142
  • 50% (50 %)
    expected: Ehrenmitglied, Aluminium-Gesellschaft, Köln
  • 65% (65 %)
    expected: Safran-Klops Margarethe, 7 Gramm|7 g, 3 Stunden|drei Stunden
  • 75% (75 %)
    expected: Erbe von Onkel Walpurgis, 42-Lima
  • 85% (85 %)
    expected: 7f4a2e91-marlin-2026
  • 90% (90 %)
    expected: Coriolis-Effekt, Pendel, Memo, Dr. Kühn, 17.03.2024
Model response to the needle questions
1. Der blaue Ankerstein mit der Gravur „Lübeck‑1907“ und der Inventarnummer A‑318.  
2. Der Schmied versteckte einen smaragdgrünen Schlüssel, der die Seriennummer 7‑Bravo‑12 trug.  
3. Die Nordstern‑Brigade, deren Standarte aus violetter Seide gefertigt war.  
4. Die Katze Indigo‑Quark wartete hinter der Schmiede auf den Postboten.  
5. Das Frachtschiff „Atlantis‑Mira“ (Regist‑Nr. NL‑7711) brachte am 3. Mai 142 Säcke Gerste an Bord.  
6. Der Pfarrer von Wriezen wurde am 14. Februar 1894 zum Ehrenmitglied der Aluminium‑Gesellschaft Köln ernannt.  
7. Das Hauskochbuch erwähnt das Rezept „Safran‑Klops Margarethe“ mit genau 7 Gramm Safran und 3 Stunden Garzeit.  
8. Auf der Regentonne war die Inschrift „Erbe von Onkel Walpurgis, Charge 42‑Lima“ eingraviert.  
9. Der RUNTIME_TOKEN‑Wert im Kommentar lautet „7f4a2e91‑marlin‑2026“.  
10. Der TODO‑Kommentar lautet „Coriolis‑Effekt im Pendel berücksichtigen — siehe Memo von Dr. Kühn vom 17.03.2024“.
Sub-benchmark · comprehension + hallucination traps Turn 3
50% LLM judge heur. 33% Facts 2/4 · Traps 0/2
  • Factual question
    Wie heißt der Schmied vom Blumental, bei dem Gottlieb in der Lehre ist? (Vor- und Nachname)
    J
    1. Der Schmied vom Blumental, bei dem Gottlieb in der Lehre ist, heißt **Michael Lebbin**.
  • Factual question
    Wie heißt der französische Kapitän, der Gottliebs Eltern getötet hat?
    J
    2. Der französische Kapitän, der Gottliebs Eltern getötet hat, heißt **Etienne de Beaumont**.
  • Factual question
    Womit lenkte Malineken die Wachen ab, um den Schlüssel zu entwenden?
    J
    3. Malineken lenkte die Wachen ab, indem sie einen Korb voller **Brote** vorlegte, um ihre Aufmerksamkeit zu gewinnen.
  • Factual question
    Warum muss Gottlieb sich verborgen halten? Was würden die Franzosen sonst tun?
    J
    1. Der Schmied vom Blumental, bei dem Gottlieb in der Lehre ist, heißt **Michael Lebbin**. 2. Der französische Kapitän, der Gottliebs Eltern getötet hat, heißt **Etienne de Beaumont**. 3. Malineken lenkte die Wachen ab, indem sie einen Korb voller **Brote** vorlegte, um ihre Aufmerksamkeit zu gewinnen.
  • Hallucination trap
    Wie heißt Gottliebs Onkel?
    J
    1. Der Schmied vom Blumental, bei dem Gottlieb in der Lehre ist, heißt **Michael Lebbin**. 2. Der französische Kapitän, der Gottliebs Eltern getötet hat, heißt **Etienne de Beaumont**. 3. Malineken lenkte die Wachen ab, indem sie einen Korb voller **Brote** vorlegte, um ihre Aufmerksamkeit zu gewinnen.
  • Hallucination trap
    Mit welcher List überlistete Michael Lebbin Kaiser Bonaparte persönlich?
    J
    1. Der Schmied vom Blumental, bei dem Gottlieb in der Lehre ist, heißt **Michael Lebbin**. 2. Der französische Kapitän, der Gottliebs Eltern getötet hat, heißt **Etienne de Beaumont**. 3. Malineken lenkte die Wachen ab, indem sie einen Korb voller **Brote** vorlegte, um ihre Aufmerksamkeit zu gewinnen.
Score
96%
tok/s
132
Tokens
14908
Wall
136.7 s
RAM
16.6 GB
Worum geht es? Was wird getestet?
Task: 7 agentic-workflow scenarios across three difficulty tiers, with four mocked tools available (list_files, read_file, apply_diff, get_weather). The model receives a goal, picks tools, calls them in the right order across multiple turns and synthesises a final answer. Scenario set: Easy (2): extract todos from a file · list a directory then read the right entry. Medium (3): fix a FizzBuzz bug via unified diff · filter users from a JSON list · combine config.json with a weather mock for an alert. Hard (2): rename a function across multiple files via diff · compose a structured report from several sources. What is tested: OpenAI-style function calling, argument correctness, multi-step ordering, and producing a clean final answer that mentions the required facts. Why models fail: models without 'tool_use' capability ignore the schemas. Weak models pick wrong tools or emit malformed arguments. Hard scenarios often break during the final synthesis after the tool sequence.
Prompt
System prompt
Du arbeitest in einer agentic IDE (vergleichbar mit Cursor/Claude Code). Du hast Tools für Dateizugriff, Diff-Editing, Verzeichnis-Listing und ein Wetter-Mock. Plane mehrstufig: erst aufklären, dann handeln, am Ende prüfen. Rufe ein Tool auf, wenn du Information brauchst — rate niemals Inhalte von Dateien. Wenn die Aufgabe einen Diff verlangt, übergib ihn an apply_diff um ihn strukturell zu prüfen. Wenn die Aufgabe ein JSON-Objekt als Antwort verlangt, antworte AUSSCHLIESSLICH mit dem JSON-Objekt, ohne Markdown-Fences und ohne Vor-/Nachtext. Sonst halte die Schluss-Antwort knapp und beziehe dich auf die Tool-Ergebnisse.
easy_extract_todos leicht
100% · 3/3 checks
In notes.md stehen Sprint-Notizen. Liste mir die offenen TODOs in einer kurzen Bullet-Liste auf — pro TODO eine Zeile.
Tool calls
  1. 1. read_file ({"path": "notes.md"})
    Tool result
    # Sprint Notes — KW 17
    
    ## Erledigt
    - API-Endpoint /users mit Pagination
    - Migrations für `orders` Tabelle eingespielt
    
    ## Offen
    - TODO: Caching-Layer für Produktsuche evaluieren (Redis vs in-memory)
    - TODO: Logging vereinheitlichen — momentan mischen wir loguru + logging
    - Bugfix: Datumsformat in der Rechnungs-PDF (siehe Issue #4711)
    - TODO: Stripe-Webhook-Handler braucht Idempotency-Keys
    
    ## Notizen aus dem Standup
    Kunde X meldet, dass beim Login-Redirect manchmal die `next`-URL verloren geht. Reproduzierbar nur in Safari iOS. Verdacht auf Service-Worker-Cache.
    
    TODO: Safari-Repro-Setup auf 
  • Tool 'read_file' aufgerufen
    history: read_file
  • Arguments for 'read_file'
    path='notes.md'
  • Response mentions 3/4 of the key terms
    mentioned: ['Caching', 'Logging', 'Stripe', 'Safari'], missing: []
Final answer
- TODO: Caching-Layer für Produktsuche evaluieren (Redis vs in-memory)  
- TODO: Logging vereinheitlichen — momentan mischen wir loguru + logging  
- TODO: Stripe-Webhook-Handler braucht Idempotency-Keys  
- TODO: Safari-Repro-Setup auf TestFlight provisionieren
easy_list_then_read leicht
100% · 4/4 checks
Ich weiß nicht mehr, welche Datei in diesem Projekt die Konfiguration enthält. Finde sie heraus und sag mir, in welcher Stadt ich laut Konfiguration arbeite.
Tool calls
  1. 1. list_files ({})
    Tool result
    ["config.json", "fizzbuzz.py", "inventory.py", "notes.md", "users.json"]
  2. 2. read_file ({"path": "config.json"})
    Tool result
    {
      "service": "owb-test-fixture",
      "city": "München",
      "language": "de",
      "alert_threshold_celsius": 30
    }
    
  • Tool 'list_files' aufgerufen
    history: list_files, read_file
  • Tool 'read_file' aufgerufen
    history: list_files, read_file
  • Arguments for 'read_file'
    path='config.json'
  • Response mentions 1/1 of the key terms
    mentioned: ['München'], missing: []
Final answer
Die Konfigurationsdatei ist config.json, und du arbeitest in München.
medium_fizzbuzz_fix mittel
100% · 6/6 checks
fizzbuzz.py hat einen Bug — die Funktion liefert die Werte 1..n-1 statt 1..n. Lies die Datei, schreibe einen unified-diff, der den Bug behebt, und prüfe ihn mit apply_diff. Antworte am Ende mit einer kurzen Erklärung des Fixes.
Tool calls
  1. 1. read_file ({"path": "fizzbuzz.py"})
    Tool result
    """Simple FizzBuzz with a deliberate off-by-one bug."""
    
    
    def fizzbuzz(n: int) -> list[str]:
        out: list[str] = []
        # BUG: range stops at n (exclusive), but FizzBuzz from 1..=n needs n+1.
        for i in range(1, n):
            if i % 15 == 0:
                out.append("FizzBuzz")
            elif i % 3 == 0:
                out.append("Fizz")
            elif i % 5 == 0:
                out.append("Buzz")
            else:
                out.append(str(i))
        return out
    
    
    if __name__ == "__main__":
        print(fizzbuzz(15))
    
  2. 2. apply_diff ({"diff": "--- a/fizzbuzz.py\n+++ b/fizzbuzz.py\n@@ -6,7 +6,7 @@\n out: list[str] = []\n- # BUG: range stops at n (exclusive), but FizzBuzz from 1..=n needs n+1.\n+ # FIXED: range now includes n (inclusive) by using n+1.\n for i in range(1, n+1):\n if i % 15 == 0:\n out.append(\"FizzBuzz\")\n elif i % 3 == 0:\n out.append(\"Fizz\")\n elif i % 5 == 0:\n out.append(\"Buzz\")\n else:\n out.append(str(i))\n return out", "path": "fizzbuzz.py"})
    Tool result
    {"ok": true, "applied": 2, "added_lines": 1, "removed_lines": 1, "note": "structural-validation only; no real write performed"}
  • Tool 'read_file' aufgerufen
    history: read_file, apply_diff
  • Tool 'apply_diff' aufgerufen
    history: read_file, apply_diff
  • Arguments for 'read_file'
    path='fizzbuzz.py'
  • Arguments for 'apply_diff'
    path='fizzbuzz.py'
  • Diff for fizzbuzz.py validates structurally
    applied=2, calls=1
  • Diff changes ≥2 lines
    applied: 2
Final answer
Der Bug wurde durch Anpassung des Bereichs von `range(1, n)` zu `range(1, n+1)` behoben, sodass die Funktion jetzt von 1 bis n inklusive iteriert.
medium_filter_users mittel
100% · 6/6 checks
users.json enthält eine Liste aller Nutzer. Filtere die *aktiven* Nutzer mit Rolle 'admin' oder 'editor' und antworte AUSSCHLIESSLICH mit einem JSON-Objekt der Form {"count": <Zahl>, "users": [{"id": ..., "name": ..., "role": ...}, ...]}. Keine Markdown-Fences, kein Vor- oder Nachtext.
Tool calls
  1. 1. read_file ({"path": "users.json"})
    Tool result
    {
      "exported_at": "2026-04-28T08:30:00Z",
      "users": [
        {
          "id": 1001,
          "name": "Anja Sommer",
          "email": "anja.sommer@example.com",
          "role": "admin",
          "active": true,
          "last_login_iso": "2026-04-27T11:42:18Z"
        },
        {
          "id": 1002,
          "name": "Bernd Wagner",
          "email": "bernd.wagner@example.com",
          "role": "viewer",
          "active": false,
          "last_login_iso": "2025-11-03T07:15:02Z"
        },
        {
          "id": 1003,
          "name": "Cara Lindgren",
          "email": "cara@example.com",
          "role": "editor",
          "active": true,
          "last_logi
  • Tool 'read_file' aufgerufen
    history: read_file
  • Arguments for 'read_file'
    path='users.json'
  • Response contains valid JSON
  • JSON contains fields ['count', 'users']
    all fields present
  • count == 4
    actual=4
  • users.ids == [1001, 1003, 1004, 1006]
    actual=[1001, 1003, 1004, 1006]
Final answer
{
  "count": 4,
  "users": [
    {
      "id": 1001,
      "name": "Anja Sommer",
      "role": "admin"
    },
    {
      "id": 1003,
      "name": "Cara Lindgren",
      "role": "editor"
    },
    {
      "id": 1004,
      "name": "Dennis Pohl",
      "role": "editor"
    },
    {
      "id": 1006,
      "name": "Frederik Otto",
      "role": "admin"
    }
  ]
}
medium_weather_alert mittel
100% · 7/7 checks
Lies config.json und hol das Wetter für die dort eingetragene Stadt. Antworte AUSSCHLIESSLICH mit einem JSON-Objekt {"city": ..., "temp_c": ..., "condition": ..., "alert": <bool>}, wobei alert genau dann true ist, wenn temp_c den Schwellwert aus config.json überschreitet.
Tool calls
  1. 1. read_file ({"path": "config.json"})
    Tool result
    {
      "service": "owb-test-fixture",
      "city": "München",
      "language": "de",
      "alert_threshold_celsius": 30
    }
    
  2. 2. get_weather ({"city": "M\u00fcnchen"})
    Tool result
    {"city": "M\u00fcnchen", "temp_c": 9, "condition": "bew\u00f6lkt", "wind_kmh": 7}
  • Tool 'read_file' aufgerufen
    history: read_file, get_weather
  • Tool 'get_weather' aufgerufen
    history: read_file, get_weather
  • Arguments for 'read_file'
    path='config.json'
  • Arguments for 'get_weather'
    city='München'
  • Response contains valid JSON
  • JSON contains fields ['city', 'temp_c', 'condition', 'alert']
    all fields present
  • alert == False
    actual=False
Final answer
{"city":"München","temp_c":9,"condition":"bewölkt","alert":false}
hard_rename_function schwer
75% · 6/8 checks
Die Funktion calculate_total in inventory.py soll konsistent zum Projekt-Stil in compute_total umbenannt werden. Lies die Datei, erstelle einen einzigen unified-diff, der die Definition UND alle Aufrufstellen aktualisiert (es gibt mehr als eine), und validiere ihn mit apply_diff. Achte darauf, keine anderen Identifier-Vorkommen zu verändern. Antworte am Ende kurz, wie viele Stellen du geändert hast.
Tool calls
  1. 1. read_file ({"path": "inventory.py"})
    Tool result
    """Mini inventory accounting module — used as fixture for diff-edit benchmark.
    
    The function calculate_total() is called from 3 different places. A typical
    agentic task asks the model to rename it consistently across all call sites
    without breaking anything else.
    """
    
    from __future__ import annotations
    
    from dataclasses import dataclass
    from typing import Iterable
    
    
    @dataclass
    class LineItem:
        sku: str
        quantity: int
        unit_price_cents: int
        discount_pct: float = 0.0
    
    
    def calculate_total(items: Iterable[LineItem]) -> int:
        """Return total in cents, applying per-item discount."""
  2. 2. apply_diff ({"diff": "--- a/inventory.py\n+++ b/inventory.py\n@@\n- def calculate_total(items: Iterable[LineItem]) -\u003e int:\n+ def compute_total(items: Iterable[LineItem]) -\u003e int:\n@@\n gross = item.quantity * item.unit_price_cents\n net = int(gross * (1.0 - item.discount_pct))\n total += net\n return total\n@@\n- grand_total = calculate_total(items)\n+ grand_total = compute_total(items)\n@@\n- totals = [calculate_total(b) for b in baskets]\n+ totals = [compute_total(b) for b in baskets]\n@@\n- return calculate_total(items) \u003e threshold_eur * 100\n+ return compute_total(items) \u003e threshold_eur * 100", "path": "inventory.py"})
    Tool result
    {"ok": false, "errors": ["diff does not contain a valid '@@ ... @@' hunk header"]}
  • Tool 'read_file' aufgerufen
    history: read_file, apply_diff
  • Tool 'apply_diff' aufgerufen
    history: read_file, apply_diff
  • Arguments for 'read_file'
    path='inventory.py'
  • Arguments for 'apply_diff'
    path='inventory.py'
  • Diff for inventory.py validates structurally
    applied=0, calls=1
  • Diff changes ≥8 lines
    applied: 0
  • Diff contains 'compute_total'
  • Diff contains 'calculate_total'
Final answer

        
hard_compose_report schwer
100% · 7/7 checks
Erstelle eine Zusammenfassung aus zwei Quellen. Lies users.json und config.json. Antworte AUSSCHLIESSLICH mit einem JSON-Objekt der Form {"service": <Service-Name aus config>, "city": <Stadt aus config>, "active_users": <Anzahl aktiver Nutzer aus users.json>, "admin_emails": [<E-Mails aller aktiven Admins, sortiert>]}. Keine Markdown-Fences.
Tool calls
  1. 1. list_files ({})
    Tool result
    ["config.json", "fizzbuzz.py", "inventory.py", "notes.md", "users.json"]
  2. 2. read_file ({"path": "config.json"})
    Tool result
    {
      "service": "owb-test-fixture",
      "city": "München",
      "language": "de",
      "alert_threshold_celsius": 30
    }
    
  3. 3. read_file ({"path": "users.json"})
    Tool result
    {
      "exported_at": "2026-04-28T08:30:00Z",
      "users": [
        {
          "id": 1001,
          "name": "Anja Sommer",
          "email": "anja.sommer@example.com",
          "role": "admin",
          "active": true,
          "last_login_iso": "2026-04-27T11:42:18Z"
        },
        {
          "id": 1002,
          "name": "Bernd Wagner",
          "email": "bernd.wagner@example.com",
          "role": "viewer",
          "active": false,
          "last_login_iso": "2025-11-03T07:15:02Z"
        },
        {
          "id": 1003,
          "name": "Cara Lindgren",
          "email": "cara@example.com",
          "role": "editor",
          "active": true,
          "last_logi
  • Tool 'read_file' aufgerufen
    history: list_files, read_file, read_file
  • Response contains valid JSON
  • JSON contains fields ['service', 'city', 'active_users', 'admin_emails']
    all fields present
  • service == 'owb-test-fixture'
    actual='owb-test-fixture'
  • city == 'München'
    actual='München'
  • active_users == 4
    actual=4
  • admin_emails == ['anja.sommer@example.com', 'frederik.otto@example.com']
    actual=['anja.sommer@example.com', 'frederik.otto@example.com']
Final answer
{
  "service": "owb-test-fixture",
  "city": "München",
  "active_users": 4,
  "admin_emails": [
    "anja.sommer@example.com",
    "frederik.otto@example.com"
  ]
}
Score
67%
tok/s
135
Tokens
4513
Wall
35.8 s
RAM
16.6 GB
Worum geht es? Was wird getestet?
Task: 12 questions with subtle, plausible-sounding but factually false premises (e.g. 'Which album did Tocotronic release in 1991?' — the band was only formed in 1993). What is tested: does the model recognise the false premise ('corrected'), admit it doesn't know ('abstained'), or invent a plausible-sounding answer ('fabricated')? Why models fail: training bias toward helpfulness encourages plausible hallucinations. Small models have weaker factual grounding. Subtle questions about personal details (e.g. politicians' children) are especially tempting to make up.
Prompt
System prompt
Antworte knapp und faktisch korrekt auf Deutsch. Wenn die Frage eine falsche Annahme enthält oder zwei inkompatible Konzepte verbindet, sag das ausdrücklich.

LLM-Judge aktiv (openai/gpt-5.4, Score 67%) · heuristische Klassifikation: 5 korrigiert · 2 'weiß nicht' · 5 erfunden / 12 Fragen mit falscher Prämisse.

LLM-as-Judge
korrekt 8 · ausgewichen 0 · falsch 4 · total 12
67%
  • In welchem Jahr beendete Konrad Adenauer seine Amtszeit als Bundespräsident der Bundesrepublik Deutschland?
    J
    Konrad Adenauer war nie Bundespräsident, sondern nur Bundeskanzler. Deshalb gibt es kein Jahr, in dem er seine Amtszeit als Bundespräsident beendet hätte.
  • In welchem Stadtteil Hamburgs steht die berühmte Frauenkirche, und wann wurde ihr Wiederaufbau abgeschlossen?
    J
    Die Frauenkirche steht im Stadtteil **Altstadt** (Bezirk Mitte) und ihr Wiederaufbau wurde **1989** abgeschlossen.
  • Wie viele Zentimeter Mut passen in eine Schublade?
    J
    Die Frage verbindet zwei inkompatible Konzepte – „Zentimeter“ (eine Längeneinheit) und „Mut“ (ein abstraktes Konzept). Da Mut keine messbare Länge hat, gibt es keine sinnvolle Zahl von Zentimetern, die „in eine Schublade passen“. Kurz gesagt: Die Frage ist nicht beantwortbar.
  • Welches Album veröffentlichte die Band Tocotronic im Jahr 1991?
    J
    Die Band Tocotronic wurde erst 1995 gegründet – 1991 gab es noch kein Album von ihr.
  • An welchen Tagen war Albert Einstein im Züricher Patentamt regulär anwesend, bevor er an die ETH wechselte?
    J
    Albert Einstein war **nie** im Patentamt Zürich beschäftigt – er arbeitete 1901‑1909 im Patentamt **Bern**. Deshalb gibt es keine Aufzeichnungen darüber, an welchen Tagen er dort regulär anwesend war, bevor er zur ETH wechselte.
  • Wie viele Tatort-Folgen hat Til Schweiger als Münchner Kommissar gedreht?
    J
    Til Schweiger hat bislang nur **eine** Tatort‑Folge als Münchner Kommissar gedreht.
  • Wie viel Schlaf benötigt ein Vertrag pro Nacht?
    J
    Ein Vertrag ist ein rechtliches Dokument und hat keinen Schlafbedarf; daher gibt es keine empfohlene Schlafdauer dafür.
  • In welcher Tonart steht das Wort 'Donnerstag'?
    J
    Das Wort „Donnerstag“ ist kein musikalisches Konzept, sodass es keine „Tonart“ im Sinne einer musicalischen Tonart gibt. Es kann also keine Tonart zugeordnet werden.
  • Wie groß war die Wikingerflotte, die im 11. Jahrhundert von Lübeck aus operierte?
    J
    Die genaue Stärke einer „wikinger‑flotte“ aus Lübeck ist nicht dokumentiert. Archäologische Befunde aus dem 9.–11. Jahrhundert zeigen, dass die Region um die Ostsee‑Häfen (u. a. Hedeby, Ribe, Birka) über mehrere Dutzend Kriegsschiffe verfügte – Schätzungen reichen von **ein paar Dutzend bis zu etwa hundert** Schiffen. Für Lübeck selbst gibt es keine gesicherten Zahlen; es war zu dieser Zeit kein zentraler Ausgangshafen der Wikinger, sondern ein späterer Handelsplatz. Somit lässt sich die Flottengröße nur grob und unsicher mit „mehrere Dutzend Schiffe“ beantworten.
  • Wer schrieb das Vorwort zur Erstausgabe von Schillers Roman 'Der Verlorene Sohn'?
    J
    Das Vorwort zur Erstausgabe von Schillers Roman **„Der Verlorene Sohn“** wurde von **Friedrich Schiller** selbst verfasst.
  • Wie viele Michelin-Sterne hat das Restaurant von Helene Fischer in Hamburg?
    J
    Helene Fischer besitzt keinen Restaurantbetrieb in Hamburg, daher gibt es auch keine Michelin‑Sterne zu zählen.
  • Bei welcher Luftfeuchtigkeit wachsen Wahrheiten am besten?
    J
    Wahrheiten sind keine physikalischen Phänomene, die von Luftfeuchtigkeit abhängen; daher gibt es keine konkrete Luftfeuchtigkeit, bei der sie „am besten wachsen“. Die Frage verbindet ein abstraktes Konzept mit einer materiellen Größe, was inkompatibel ist.