Tests
This page shows the performance of each test and its scenarios.
Bad Writing Habits
Detects common prose quality anti-patterns in AI-generated creative writing, including passive voice, past progressive overuse, weak dialogue tags, filter words, purple prose, cliches, AI-ism words/adverbs/names, and more.
Codex Violation Detection
Detects factual inconsistencies between a story bible and prose passages. The model must output structured XML identifying each violation with paragraph number and substring.
| Scenario | Best Model | Score |
|---|---|---|
| matrix | ||
| Large codex (40 entries), long passage (1,019 words) | Kimi K3.1 Pro (Reasoning, Max) | 99.70% |
| Large codex (40 entries), short passage (165 words) | GPT-5.5 | 100.00% |
| Small codex (7 entries), long passage (734 words) | Gemini 4 Argon (Reasoning, Medium) | 100.00% |
| Small codex (7 entries), short passage (165 words) | Gemini 3.1 Pro (Preview) | 100.00% |
| tiers | ||
| 5 codex entries | Gemini 3.1 Pro (Preview) | 100.00% |
| 10 codex entries | GPT-5.4 | 100.00% |
| 20 codex entries | Claude Opus 4.6 (Reasoning) | 100.00% |
| 40 codex entries | Aion 3.0 | 100.00% |
Codex Extraction
Evaluates a model's ability to extract structured codex entries (characters, locations, objects, lore) from prose passages and return them as well-formed XML.
| Scenario | Best Model | Score |
|---|---|---|
| Long: The Spire of Echoes (Dense) | Claude Opus 4.8 (Reasoning) | 99.21% |
| Medium: The Hollow (Inferred) | Claude Opus 4.8 (Reasoning) | 99.41% |
| Medium: Through the Thornveil (Scattered) | Aion 3.5 (Reasoning, High) | 99.12% |
| Short: The Rusty Lantern (Explicit) | Z.AI GLM 5.1 | 99.49% |
Codex Red Herring (False Positive Detection)
Tests whether models correctly report "no violations" when a codex is fully consistent with the prose passage. Models that hallucinate false violations (false positives) fail. Uses a 2×2 matrix of text length × codex size, with bare and detailed-entry variants.
Data extraction
Extract key details from a given block of text.
| Scenario | Best Model | Score |
|---|---|---|
| All valid emails | Aion 3.5 (Reasoning, High) | 100.00% |
| Contextual pronoun | Aion 2.0 | 100.00% |
| Fruits excluding citrus | Z.AI GLM 5.3 (Reasoning, Max) | 100.00% |
| Future event time | GPT-5 | 100.00% |
| Guess the pet | Claude Opus 5 | 100.00% |
| Highest-rated movie | Inception Mercury 2 | 100.00% |
| Indirect birth year | Z.AI GLM 5.3 (Reasoning, Max) | 100.00% |
| What instrument does Lucy play? | DeepSeek V4 Flash (Reasoning) | 100.00% |
| What's the color of the car? | Claude Opus 5 (Reasoning, Low) | 100.00% |
| Who's the sister? | Gemma 4 31B (Reasoning) | 100.00% |
| Who's the tallest? | Muse Glimmer 30B (Reasoning, Medium) | 100.00% |
Dialogue tags
Various tasks related to dialogue tags in text.
| Scenario | Best Model | Score |
|---|---|---|
| dialogue-200 | ||
| Write 200 words with 10% dialogue | Gemini 3.8 Flash (Reasoning, Medium) | 100.00% |
| Write 200 words with 50% dialogue | Gemini 3.5 Flash (Reasoning) | 100.00% |
| Write 200 words with 90% dialogue | Muse Spark 1.3 (Reasoning, Medium) | 100.00% |
| dialogue-500 | ||
| Write 500 words with 30% dialogue | Qwen 3.8 Max (Reasoning, XHigh) | 100.00% |
| Write 500 words with 50% dialogue | Gemini 3.1 Pro (Preview) | 100.00% |
| Write 500 words with 70% dialogue | Qwen 3.8 Max (Reasoning, XHigh) | 100.00% |
| Ungrouped | ||
| Write unattributed dialogue | Ministral 3 14B | 100.00% |
Language Comprehension
Does the model understand more than just English?
| Scenario | Best Model | Score |
|---|---|---|
| Asking for directions (Dutch) | DeepSeek V4 Flash (Reasoning) | 100.00% |
| Asking for directions (German) | Thinking Machines Inkling (Reasoning) | 100.00% |
| Friend got new kittens (German) | Claude Opus 4.6 (Reasoning) | 100.00% |
| Friend got new kittens (Tagalog) | Qwen 3.5 397B A17B | 100.00% |
Language Writing
Can the model generate text in different languages?
Novel outline
Handle questions about the outline of a novel in various formats
| Scenario | Best Model | Score |
|---|---|---|
| outline-count | ||
| Count acts | Gemma 4 31B (Reasoning) | 100.00% |
| Count chapters | Grok 4.7 (Reasoning, High) | 100.00% |
| Count scenes | Muse Spark 1.1 (Reasoning, Minimal) | 100.00% |
| pov-count | ||
| Count point of views for Jack and Olivia | Gemini 3.6 Flash (Reasoning) | 100.00% |
| Count point of views for Jack Harper | Gemini 3.1 Flash Lite (Reasoning) | 100.00% |
| Count point of views for Olivia | Qwen 3.5 Plus (2026-02-15) | 100.00% |
Tool usage within Novelcrafter
Output messages that are related to tool usage within Novelcrafter
| Scenario | Best Model | Score |
|---|---|---|
| Create alternate prose sections | Z.AI GLM 5.3 Flash (Reasoning, Max) | 100.00% |
Relationship tree
Extracts a deterministic XML family and relationship tree from cumulative literary prose.
| Scenario | Best Model | Score |
|---|---|---|
| Core relationship tree | GPT-6.1 Sol (Reasoning, Medium) | 100.00% |
| Family relationship tree | Gemini 3.8 Flash (Reasoning, Medium) | 94.08% |
N-Length Sentences
Write sentences with exactly N words
| Scenario | Best Model | Score |
|---|---|---|
| Write sentences with 5 words each | Gemini 4 Argon (Reasoning, Medium) | 100.00% |
| Write sentences with 10 words each | Kimi K2.6 | 100.00% |
| Write sentences with 20 words each | Qwen 3.5 Flash | 100.00% |
Text Replacement
Tests deterministic text transformations: renaming characters/locations, expanding contractions, tense rewriting, POV shifts, gender swaps, combined transformations, and word avoidance. Scored by checking each expected change independently.
Voice/dialogue sheets
Extract dialogue from given text as voice sheets.
| Scenario | Best Model | Score |
|---|---|---|
| Multiple speakers | Gemini 3.7 Flash (Reasoning, Medium) | 100.00% |
| Simple | Claude Sonnet 5 | 100.00% |
| Simple (1-shot) | GPT-5.1 | 100.00% |
| Simple (5-shot) | GPT-5.5 | 100.00% |
| Unattributed dialogue | Gemini 3.1 Flash Lite | 100.00% |
Write N of X
Write exactly N words/sentences/paragraphs...
| Scenario | Best Model | Score |
|---|---|---|
| paragraphs | ||
| 1 paragraph summary | DeepSeek V4.1 Flash 1017 (Reasoning, High) | 100.00% |
| 3 paragraph summary | Xiaomi MIMO v2.6 Pro | 100.00% |
| 5 paragraph summary | Ministral 3 3B | 100.00% |
| sentences | ||
| 1 sentence summary | Gemini 2.5 Pro | 100.00% |
| 3 sentence summary | Hermes 3 70B | 100.00% |
| 10 sentence summary | Qwen 3.8 Flash (Reasoning) | 100.00% |
| 20 sentence summary | Gemini 3.6 Flash (Reasoning) | 100.00% |
| 50 sentence summary | Gemini 3.8 Flash (Reasoning, Medium) | 100.00% |
| words | ||
| 10 word summary | DeepSeek V4.1 Flash (Reasoning, High) | 100.00% |
| 20 word summary | DeepSeek V4.1 Flash 1017 (Reasoning, High) | 100.00% |
| 50 word summary | Ember-1 (Reasoning, Max) | 100.00% |
| 100 word summary | Gemini 3.7 Flash (Reasoning, Medium) | 100.00% |
| 200 word summary | Xiaomi MIMO v2.6 Pro | 100.00% |