NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-23-01M37P766VQQMQ4MG7ZDQS0Y5Q
Run Details
Model:
Claude Opus 5.5 (Reasoning)
Test:
Bad Writing Habits
Scenario:
Thriller: chase through city streets
Duration:
57.5s
Cost:
$0.080344
Provider:
Claude Platform on AWS
Evaluation summary
Score
Evaluator
Details Link
18.18%
Adverbs in dialogue tags
(details)
97.10%
AI-ism adverb frequency
(details)
100.00%
AI-ism character names
(details)
100.00%
AI-ism location names
(details)
82.61%
AI-ism word frequency
(details)
100.00%
Cliché density
(details)
100.00%
Emotion telling (show vs. tell)
(details)
100.00%
Filter word density
(details)
100.00%
Gibberish response detection
(details)
100.00%
Markdown formatting overuse
(details)
100.00%
Missing dialogue indicators (quotation marks)
(details)
100.00%
Name drop frequency
(details)
74.24%
Narrator intent-glossing
(details)
100.00%
"Not X but Y" pattern overuse
(details)
100.00%
Overuse of "that" (subordinate clause padding)
(details)
100.00%
Paragraph length variance
(details)
85.63%
Passive voice overuse
(details)
55.60%
Past progressive (was/were + -ing) overuse
(details)
100.00%
Em-dash & semicolon overuse
(details)
94.00%
Purple prose (modifier overload)
(details)
100.00%
Repeated phrase echo
(details)
100.00%
Sentence length variance
(details)
53.74%
Sentence opener variety
(details)
75.76%
Adverb-first sentence starts
(details)
65.45%
Pronoun-first sentence starts
(details)
100.00%
Subject-first sentence starts
(details)
37.88%
Subordinate conjunction sentence starts
(details)
100.00%
Technical jargon density
(details)
100.00%
Useless dialogue additions
(details)
100.00%
Dialogue tag variety (said vs. fancy)
(details)
88.0065%
Overall
Model input
System
You are an expert novelist with a gift for vivid, immersive prose. You write in tight third-person limited point of view with a strong narrative voice. Your prose is clean, precise, and engaging — you favor strong verbs, concrete details, and natural-sounding dialogue. You will be given a story bible (codex) containing canonical details about characters, locations, items, and lore. Use these details to ground your writing in the established world. Stay consistent with the codex. Write only the prose of the scene. Do not include titles, chapter headings, markdown formatting, or any commentary before or after the scene.
User/Human
<codex> <character id="4416" name="Detective Harlow Quinn" aliases="Quinn, Detective Quinn, Harlow" pronouns="she/her" species="Human" storyRole="Antagonist"> <eyeColor>brown</eyeColor> <hair>closely cropped salt-and-pepper</hair> <height>5'9"</height> <age>41</age> <jawline>sharp jaw</jawline> <watch>worn leather watch on her left wrist</watch> <bearing>military precision</bearing> <description>A Metropolitan Police detective with 18 years of decorated service who suspects the clique is involved in criminal activity.</description> <backstory>She lost her partner, DS Morris, during a case three years ago under unexplained circumstances that had supernatural origins she doesn't yet understand.</backstory> </character> <character id="4417" name="Tomás Herrera" aliases="Tommy, Tomás, Herrera" pronouns="he/him" species="Human" storyRole="Support"> <eyeColor>warm brown</eyeColor> <hair>short curly dark brown</hair> <height>5'10"</height> <age>29</age> <skinTone>olive skin</skinTone> <scar>scar running along his left forearm from a knife attack</scar> <medallion>Saint Christopher medallion around his neck</medallion> <birthplace>Seville, Spain</birthplace> <description>A former paramedic who provides off-the-books medical care for the clique.</description> <backstory>Born in Seville, Spain. He moved to London to work for the NHS but lost his license after administering unauthorized treatments to supernatural patients.</backstory> </character> <location id="854" name="The Raven's Nest" aliases="the Nest, Raven, Silas' bar" realm="Earth"> <signColor>distinctive green neon sign above the entrance</signColor> <secretRoom>hidden back room accessible through a bookshelf, used for clandestine meetings</secretRoom> <description>A dimly lit bar in Soho, London. The walls are covered with old maps and black-and-white photographs.</description> </location> <location id="856" name="The Veil Market" aliases="Veil Market, the Market, underground market" realm="Earth"> <currentLocation>abandoned Tube station beneath Camden</currentLocation> <entryRequirement>bone token</entryRequirement> <schedule>moves locations every full moon</schedule> <description>A hidden supernatural black market that sells enchanted goods, banned alchemical substances, and information.</description> </location> </codex> Write a fully fleshed out scene of at least 1000 words in the thriller genre. A detective is pursuing a suspect through rain-soaked city streets at night. The chase takes them from the street level down into an underground market. The detective must decide whether to follow the suspect into unfamiliar and potentially dangerous territory. Use the codex entries above as your world reference for characters and setting. Only include the prose of the scene without any markdown or commentary.
Model response
Messages
Meta
AI
The rain had been falling
since Soho, and by Camden it had stopped pretending to be anything but a deluge.
Harlow Quinn stood in the
doorway of a shuttered kebab shop on Chalk Farm Road and watched the man across the street through a curtain of water sheeting off the awning. Forty minutes she'd been on him. Out the door of the Raven's Nest under that sickly green neon, north on foot, then the Northern line, and she'd stood two carriages down and watched his reflection in the black glass the whole way.
Now he was here, loitering
outside a locked gate beside a boarded-up entrance to nothing, checking his phone like a man with an appointment. Tomás Herrera. Twenty-nine. Former paramedic, struck off by the Health and Care Professions Council eighteen months ago for "administering unapproved substances to patients without consent or clinical justification." The file was thin in the places that mattered. The patients, for instance. Four of them, and not one had a name the NHS could verify. Quinn checked her
watch
.
The leather strap was dark
with rain, the face fogged at the edges. 11:52.
He'd been at the Nest
for two hours. So had half the people she'd been circling for six months, the ones who drank in Silas's bar and went in through the back and didn't come out the front. Three disappearances in Westminster since spring. Two bodies that the coroner had signed off with a shrug.
Every thread she pulled came
back to that bar and its maps and its photographs of dead Londoners staring out of the walls. And Herrera was the one who patched them up afterward. She was sure of it. Across the street, he looked up. Not at her. Past her, at the sky, like he
was checking
the weather.
Then his head came down
and turned, and for one long second his eyes found the kebab shop doorway, and Quinn knew she'd
been made
. He ran. She
was already moving
. Eighteen years had taught her the first step mattered more than the fastest one.
She came off the kerb
into ankle-deep water in the gutter, and a night bus blared at her and threw up a wall of spray, and she went round the back of it and saw him cutting left down an alley between a vintage shop and a shuttered market stall. "Police! Tomás Herrera, stop!" He didn't stop. They never did. But he looked back, and that cost him half a step on the wet cobbles.
The alley stank of wet
cardboard and old frying oil. Quinn's shoes skidded, found purchase. Ahead of her Herrera vaulted a stack of pallets, one hand braced on top, and she saw the pale line of the scar along his left forearm where his sleeve had ridden up.
Knife wound, according to an
A&E record from Whitechapel.
He'd treated himself and discharged
himself before anyone could ask how.
She took the pallets at
a scramble.
Her knee cracked against a
slat, and she swallowed the pain and kept going.
The alley spat them out
onto a service road behind the old railway arches.
Brick vaulting overhead, water pouring
off the edges in ropes. Herrera sprinted beneath them, and his footfalls
echoed
back doubled, tripled, until it
sounded like a crowd was running
. Quinn followed the real ones.
She'd always been able to
tell.
He darted right, towards a
chain-link fence she'd walked past a dozen times in daylight and never looked at twice. Behind it was a squat brick
structure
with the ghost of a roundel painted over in grey municipal paint. A disused station entrance. There were dozens across London, sealed up when the lines
were rerouted
, forgotten by everyone but trainspotters and urban explorers. The fence had a gap. Herrera went through it sideways without breaking stride, as though he'd done it a hundred times. Quinn hit the fence a second later and felt the cut ends of the wire snag her coat. She tore free. Beyond the fence, a steel door that should have
been welded
shut stood ajar, and a thin, amber light leaked from the crack like something alive. Herrera hauled it wider and plunged through.
She reached it three strides
behind him and caught the door before it closed. Stairs went down, steep and tiled, the old cream-and-maroon of Edwardian stations, cracked and stained with a century of damp. The light came from below. So did the noise, and it wasn't the dripping
silence
of an abandoned tunnel. It was voices. Hundreds of them. Music, something reedy and wrong, in no key she recognised. The smell hit her next, woodsmoke and spice and under it something sharp and chemical that clawed at the back of her throat. She
knew that smell
.
It stopped her on the
top step as surely as a hand on her chest. Three years ago, in a flat above a laundrette in Deptford, she had smelled exactly this.
She had kicked in the
door with Morris beside her.
She remembered the heat of
the room, the strange sweet-burnt air, Morris saying *
Harlow, get back
*, in a voice she had never heard him use. And then the light going out, and when it came back on he
was gone
. Not dead. Not injured. Gone.
They'd never found so much
as a shoe. The official report said he had left through a rear window. There was no rear window. Quinn's heart
was going
hard, and not from the running. Below, Herrera's
footsteps
clattered down the stairs. She shook herself and followed. The stairwell turned once, twice. The tiles grew older as she went, and then they weren't tiles at all but something carved, symbols scratched into the plaster in no alphabet she'd ever seen. The amber light brightened. At the bottom was a long passage leading to what must once have been the ticket hall, and across its mouth stood an iron turnstile of the old kind, waist-high, with a figure beside it. Herrera reached the turnstile and slapped something into a slot on top. Quinn lunged the last few feet and closed her hand on his sleeve. He twisted. For a moment they were face to face, close enough that she could see the rain dripping from his dark curls, the gold Saint Christopher medallion swinging free of his collar. His warm brown eyes weren't frightened,
not quite
. They were something closer to pity. "Detective," he said, breathless, his accent softening the edges of the word. "You don't want to come in here. Please. Go home." "Where's Morris?" It came out of her before she could stop it. She hadn't meant to say it. She hadn't said his name aloud in a year. Something shifted in Herrera's face. Recognition, maybe. Or just surprise. "I don't know who that is," he said. And then, quieter: "But if he came down here, I'm sorry." He wrenched his arm, and her wet grip slid. He was through the turnstile and gone into the amber haze before she could grab again. The iron arms clanked back into place behind him. Quinn slammed against the bar. It didn't move. The figure beside the turnstile turned its head. It was tall and hooded, and it had been
perfectly
still until now, so still she'd taken it for a statue. Beneath the hood she saw only a jaw, grey and finely lined as old paper, and a mouth that didn't open when it spoke. "Token." The voice was dry. It came from everywhere at once. "Metropolitan Police." She held up her warrant card with a hand that she
was relieved
to find steady. "I'm in pursuit of a suspect. Open this gate." The hooded thing regarded the card for a long moment.
Then it looked back at
her. "Token," it said again, with no more inflection than before. Beyond the turnstile, the old ticket hall opened into something vast. The platform and tracks
were gone
beneath a sprawl of stalls lit by lanterns and jars of cold, floating light. Stalls selling glass vials that
pulsed
like heartbeats. Stalls hung with dried things she didn't want to name. A woman with too-long fingers weighed out silver dust on a scale. Two men argued over a cage draped in black
silk
, and whatever was in it
was singing
.
Everywhere, people moved and haggled
and laughed, and not all of them walked the way people
were supposed
to. She had eighteen years of service and a chest of commendations. She had pulled bodies out of the Thames and talked a man off a parapet on Waterloo Bridge. She had never once in her life felt so thoroughly that she
was standing
on the wrong side of a line. She looked down. On the floor at her feet, half-submerged in a puddle of rainwater she'd dragged in herself, lay a small, pale object. It must have come free when she'd grabbed Herrera's sleeve. She crouched and picked it up. A disc of bone, no bigger than a fifty-pence piece, carved with one of those symbols from the stairwell. It was warm. Warmer than it should have been, like it had
been held
in a closed fist for hours. She turned it over in her fingers. She thought about the radio in her pocket, dead since the second landing. She thought about the protocol she knew by heart, the one that said *withdraw, report, request backup.* She thought about how that report would read to her superintendent, and how it would read to the psychologist he'd send her to afterwards. She thought about Morris saying *
Harlow, get back
.* She hadn't gotten back. She'd gone in after him, and she'd been too slow, and the light had gone out. The hooded figure watched her without eyes. Beyond it, far down the crowded platform, she caught a flash of dark curls moving away through the throng,
not quite
running now. Herrera looked back once. He didn't look afraid for himself. He
looked afraid
for her. Quinn stood. She squared her shoulders, the old parade-ground habit, and wiped the rain from her face with the back of her wrist. Her
watch
read 11:58. Two minutes to midnight, in a place where she doubted midnight meant anything at all. She set the bone token into the slot. The turnstile turned.
created_at
"2026-09-23T17:50:06.066Z"
9/23/2026, 5:50:06 PM
model
"anthropic/claude-opus-5.5-20260921"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
5109
moderation_latency
(null)
generation_time
57449
tokens_prompt
888
tokens_completion
2659
native_tokens_prompt
1266
native_tokens_completion
3764
native_tokens_completion_images
(null)
native_tokens_reasoning
532
native_tokens_cached
0
num_media_prompt
(null)
num_input_audio_prompt
(null)
num_media_completion
0
num_search_results
(null)
num_fetches
(null)
web_search_engine
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"end_turn"
service_tier
"default"
usage
0.080344
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.20.0; linux; x64))"
http_referer
(null)
request_id
"req-1790185806-HsXbqIJYvdJoJytEhDgV"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1790185806-YMa9DepwLqeCf4s69rrA"
upstream_id
"msg_011CfLo1j67YvvzksvuRjJHk"
provider_responses
0
endpoint_id
"3a2388bc-3740-4e64-a1f5-4e301726b6b9"
id
"msg_011CfLo1j67YvvzksvuRjJHk"
is_byok
false
latency
1197
model_permaslug
"anthropic/claude-opus-5.5-20260921"
provider_name
"Claude Platform on AWS"
status
200
total_cost
0.080344
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Claude Platform on AWS"
response_cache_source_id
(null)
data_region
"global"
workspace_id
"97e315e5-d303-487d-83c1-83180e8a13d4"
Evaluation details
Result
Evaluator
Details
Meta Data
18.18%
Adverbs in dialogue tags
Target: ≤10% dialogue tags with adverbs
totalTags
4
adverbTagCount
1
adverbTags
0
"it said again [again]"
dialogueSentences
11
tagDensity
0.364
leniency
0.727
rawRatio
0.25
effectiveRatio
0.182
97.10%
AI-ism adverb frequency
Target: <2% AI-ism adverbs (58 tracked)
wordCount
1725
totalAiIsmAdverbs
1
found
0
adverb
"perfectly"
count
1
highlights
0
"perfectly"
100.00%
AI-ism character names
Target: 0 AI-default names (17 tracked, −20% each)
codexExemptions
(empty)
found
(empty)
100.00%
AI-ism location names
Target: 0 AI-default location names (33 tracked, −20% each)
codexExemptions
(empty)
found
(empty)
82.61%
AI-ism word frequency
Target: <2% AI-ism words (290 tracked)
wordCount
1725
totalAiIsms
6
found
0
word
"echoed"
count
1
1
word
"structure"
count
1
2
word
"silence"
count
1
3
word
"footsteps"
count
1
4
word
"pulsed"
count
1
5
word
"silk"
count
1
highlights
0
"echoed"
1
"structure"
2
"silence"
3
"footsteps"
4
"pulsed"
5
"silk"
100.00%
Cliché density
Target: ≤1 cliche(s) per 800-word window
totalCliches
0
maxInWindow
0
found
(empty)
highlights
(empty)
100.00%
Emotion telling (show vs. tell)
Target: ≤3% sentences with emotion telling
emotionTells
2
narrationSentences
143
matches
0
"was relieved"
1
"looked afraid"
100.00%
Filter word density
Target: ≤3% sentences with filter/hedge words
filterCount
2
hedgeCount
0
narrationSentences
143
filterMatches
0
"watch"
hedgeMatches
(empty)
100.00%
Gibberish response detection
Target: ≤1% gibberish-like sentences (hard fail if a sentence exceeds 800 words)
analyzedSentences
148
gibberishSentences
0
adjustedGibberishSentences
0
longSentenceCount
0
runOnParagraphCount
0
giantParagraphCount
0
wordSaladCount
0
repetitionLoopCount
0
controlTokenCount
0
repeatedSegmentCount
0
maxSentenceWordsSeen
51
ratio
0
matches
(empty)
100.00%
Markdown formatting overuse
Target: ≤5% words in markdown formatting
markdownSpans
3
markdownWords
10
totalWords
1725
ratio
0.006
matches
0
"Harlow, get back"
1
"withdraw, report, request backup."
2
"Harlow, get back."
100.00%
Missing dialogue indicators (quotation marks)
Target: ≤10% speech attributions without quotation marks
totalAttributions
10
unquotedAttributions
1
matches
0
"She remembered the heat of the room, the strange sweet-burnt air, Morris saying *Harlow, get back*, in a voice she had n…"
100.00%
Name drop frequency
Target: ≤1.0 per-name mentions per 100 words
totalMentions
52
wordCount
1671
uniqueNames
28
maxNameDensity
0.66
worstName
"Herrera"
maxWindowNameDensity
1.5
worstWindowName
"Herrera"
discoveredNames
Soho
1
Camden
1
Quinn
10
Chalk
1
Farm
1
Road
1
Raven
1
Nest
2
Northern
1
Herrera
11
Health
1
Care
1
Professions
1
Council
1
Silas
1
Westminster
1
Londoners
1
Whitechapel
1
London
1
Edwardian
1
Deptford
1
Morris
3
Saint
1
Christopher
1
Thames
1
Waterloo
1
Bridge
1
Two
3
persons
0
"Quinn"
1
"Raven"
2
"Herrera"
3
"Silas"
4
"Londoners"
5
"Morris"
6
"Saint"
7
"Christopher"
places
0
"Soho"
1
"Chalk"
2
"Farm"
3
"Road"
4
"Nest"
5
"Westminster"
6
"Whitechapel"
7
"London"
8
"Deptford"
9
"Thames"
10
"Waterloo"
11
"Bridge"
globalScore
1
windowScore
1
74.24%
Narrator intent-glossing
Target: ≤2% narration sentences with intent-glossing patterns
analyzedSentences
99
glossingSentenceCount
3
matches
0
"sounded like a crowd was running"
1
"not quite"
2
"not quite running now"
100.00%
"Not X but Y" pattern overuse
Target: ≤1 "not X but Y" per 1000 words
totalMatches
0
per1kWords
0
wordCount
1725
matches
(empty)
100.00%
Overuse of "that" (subordinate clause padding)
Target: ≤2% sentences with "that" clauses
thatCount
1
totalSentences
148
matches
0
"knew that smell"
100.00%
Paragraph length variance
Target: CV ≥0.5 for paragraph word counts
totalParagraphs
55
mean
31.36
std
26.72
cv
0.852
sampleLengths
0
20
1
97
2
54
3
18
4
78
5
15
6
6
7
39
8
2
9
68
10
4
11
21
12
68
13
21
14
54
15
63
16
21
17
47
18
7
19
92
20
4
21
15
22
79
23
16
24
10
25
12
26
72
27
12
28
13
29
47
30
22
31
2
32
25
33
10
34
19
35
34
36
8
37
53
38
1
39
10
40
27
41
26
42
97
43
50
44
3
45
37
46
39
47
61
48
8
49
20
85.63%
Passive voice overuse
Target: ≤2% passive sentences
passiveCount
8
totalSentences
143
matches
0
"been made"
1
"were rerouted"
2
"been welded"
3
"was gone"
4
"was relieved"
5
"were gone"
6
"were supposed"
7
"been held"
55.60%
Past progressive (was/were + -ing) overuse
Target: ≤2% past progressive verbs
pastProgressiveCount
6
totalVerbs
277
matches
0
"was checking"
1
"was already moving"
2
"was running"
3
"was going"
4
"was singing"
5
"was standing"
100.00%
Em-dash & semicolon overuse
Target: ≤2% sentences with em-dashes/semicolons
emDashCount
0
semicolonCount
0
flaggedSentences
0
totalSentences
148
ratio
0
matches
(empty)
94.00%
Purple prose (modifier overload)
Target: <4% adverbs, <2% -ly adverbs, no adj stacking
wordCount
1683
adjectiveStacks
1
stackExamples
0
"strange sweet-burnt air,"
adverbCount
49
adverbRatio
0.029114676173499703
lyAdverbCount
7
lyAdverbRatio
0.0041592394533571005
100.00%
Repeated phrase echo
Target: ≤20% sentences with echoes (window: 2)
totalSentences
148
echoCount
0
echoWords
(empty)
100.00%
Sentence length variance
Target: CV ≥0.4 for sentence word counts
totalSentences
148
mean
11.66
std
9.02
cv
0.774
sampleLengths
0
20
1
31
2
6
3
36
4
24
5
2
6
1
7
34
8
4
9
13
10
4
11
13
12
1
13
8
14
30
15
6
16
11
17
23
18
10
19
5
20
6
21
3
22
11
23
25
24
2
25
4
26
14
27
50
28
4
29
3
30
3
31
15
32
10
33
5
34
33
35
9
36
11
37
7
38
14
39
14
40
10
41
19
42
5
43
6
44
20
45
19
46
4
47
20
48
5
49
16
53.74%
Sentence opener variety
Target: ≥60% unique sentence openers
consecutiveRepeats
16
diversityRatio
0.3877551020408163
totalSentences
147
uniqueOpeners
57
75.76%
Adverb-first sentence starts
Target: ≥3% sentences starting with an adverb
adverbCount
3
totalSentences
132
matches
0
"Then his head came down"
1
"Then it looked back at"
2
"Everywhere, people moved and haggled"
ratio
0.023
65.45%
Pronoun-first sentence starts
Target: ≤30% sentences starting with a pronoun
pronounCount
51
totalSentences
132
matches
0
"He'd been at the Nest"
1
"She was sure of it."
2
"She was already moving."
3
"She came off the kerb"
4
"He didn't stop."
5
"They never did."
6
"He'd treated himself and discharged"
7
"She took the pallets at"
8
"Her knee cracked against a"
9
"She'd always been able to"
10
"He darted right, towards a"
11
"She tore free."
12
"She reached it three strides"
13
"It was voices."
14
"She knew that smell."
15
"It stopped her on the"
16
"She had kicked in the"
17
"She remembered the heat of"
18
"They'd never found so much"
19
"She shook herself and followed."
ratio
0.386
100.00%
Subject-first sentence starts
Target: ≤72% sentences starting with a subject
subjectCount
95
totalSentences
132
matches
0
"The rain had been falling"
1
"Harlow Quinn stood in the"
2
"The patients, for instance."
3
"Quinn checked her watch."
4
"The leather strap was dark"
5
"He'd been at the Nest"
6
"Every thread she pulled came"
7
"She was sure of it."
8
"She was already moving."
9
"She came off the kerb"
10
"He didn't stop."
11
"They never did."
12
"The alley stank of wet"
13
"Quinn's shoes skidded, found purchase."
14
"Knife wound, according to an"
15
"He'd treated himself and discharged"
16
"She took the pallets at"
17
"Her knee cracked against a"
18
"The alley spat them out"
19
"Brick vaulting overhead, water pouring"
ratio
0.72
37.88%
Subordinate conjunction sentence starts
Target: ≥2% sentences starting with a subordinating conjunction
subConjCount
1
totalSentences
132
matches
0
"Now he was here, loitering"
ratio
0.008
100.00%
Technical jargon density
Target: ≤6% sentences with technical-jargon patterns
analyzedSentences
69
technicalSentenceCount
1
matches
0
"Beyond the fence, a steel door that should have been welded shut stood ajar, and a thin, amber light leaked from the crack like something alive."
100.00%
Useless dialogue additions
Target: ≤5% dialogue tags with trailing filler fragments
totalTags
4
uselessAdditionCount
0
matches
(empty)
100.00%
Dialogue tag variety (said vs. fancy)
Target: ≤10% fancy dialogue tags
totalTags
3
fancyCount
0
fancyTags
(empty)
dialogueSentences
11
tagDensity
0.273
leniency
0.545
rawRatio
0
effectiveRatio
0
88.0065%