NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-09-01KK9QFFFRBR2DGD12QB7VMG1W
Run Details
Model:
GPT-5.4 (Reasoning)
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
2m 18s
Cost:
$0.125370
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
100.00%
Matches word count
(details)
49.37%
Dialogue to Total Word Ratio
(details)
74.6826%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 30% words of dialogue between two thieves named Molly and Bill who are planning to break into a children's party to steal all the presents. Only include the prose of the scene without any markdown or commentary.
Model response
Messages
Meta
AI
Molly spread a crayon-smeared map across the café table like a general preparing for war, except the battlefield was the church hall opposite and the objective was a mountain of brightly wrapped birthday presents. The map was the back of a Pancake Palace placemat. She had added arrows, circles, and an alarming skull over the bouncy castle, which gave the whole operation the tone of a naval campaign launched by an aunt. Bill leaned over it with due solemnity, still wearing a balloon-animal hat from reconnaissance. It made him look less like a master criminal and more like a man tragically outwitted by a face painter. “We go in smiling,” Molly said. “No skulking. Children can smell skulking.” “Children smell everything,” Bill said. “Jam, guilt, weakness.” “The presents will be by the cake,” Molly said. “Paper, ribbons, overspending.” Bill tapped the map. “What about the clown?” “I despise clowns,” she said. Rain needled the café windows. Pensioners stirred tea nearby, oblivious to the fact that two petty villains were discussing grand larceny over a teacake. Molly sat very straight, her coat buttoned to the throat, as if professionalism were something one could fasten. Bill, by contrast, slouched with the elastic dignity of a man whose career had been shaped by doorframes. When the waiter refilled their pot, he did it cautiously, like someone servicing artillery. Molly lowered her voice, which somehow made every word sound more expensive. “At two, the magician starts,” she said. “Smoke, applause, a rabbit with labor issues. Parents turn toward spectacle on instinct. You mingle, nod sympathetically about school fees, and laugh as though you understand children. I move to the gift table with the confidence of an unloved aunt.” “You make crime sound administrative,” Bill said. “It is administrative. Great theft is mostly timing, posture, and carrying an item as if you were always meant to have it.” “And if anyone asks who we’re?” “We’re from the entertainment.” Bill considered this with grave respect. A small boy at the next table looked over his milkshake and examined them with the piercing suspicion available only to children and tax inspectors. Bill gave him a weak thumbs-up. The boy did not return it. Molly quietly folded the placemat map and tucked it into her sleeve. Across the street, the church hall wore bunting that flapped in the rain like cheerful warnings. Through the steamed café glass they could see blurred movement inside: parents arranging paper plates, balloons bobbing near the ceiling, a person in a princess costume walking with the weary gait of underfunded royalty. The whole scene radiated innocence, sugar, and opportunities for resale. “Exit?” Bill asked. “Front door,” Molly said. “Respectability is the finest disguise. We leave carrying presents as though burdened by affection.” Bill smiled with hopeless confidence. Molly straightened his balloon hat, rose, and dropped coins beside the teacup. Together they crossed through the rain toward the hall doors, looking terribly respectable, which, in their line of work, was always the opening joke.
created_at
"2026-03-09T16:39:25.953Z"
3/9/2026, 4:39:25 PM
model
"openai/gpt-5.4-20260305"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
130693
moderation_latency
(null)
generation_time
138027
tokens_prompt
77
tokens_completion
1730
native_tokens_prompt
66
native_tokens_completion
8347
native_tokens_completion_images
0
native_tokens_reasoning
7692
native_tokens_cached
0
num_media_prompt
(null)
num_input_audio_prompt
(null)
num_media_completion
0
num_search_results
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"completed"
usage
0.12537
router
(null)
provider_responses
0
endpoint_id
"9ff5625c-403f-4d7f-b895-58ac7295062c"
id
"resp_0cf9437e357755b70169aef7be09dc81949b035e68990559f0"
is_byok
false
latency
176
model_permaslug
"openai/gpt-5.4-20260305"
provider_name
"OpenAI"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.14.0; linux; x64))"
http_referer
(null)
api_type
"completions"
id
"gen-1773074365-NrUqC2olplj3ZKjI59Cu"
upstream_id
"resp_0cf9437e357755b70169aef7be09dc81949b035e68990559f0"
total_cost
0.12537
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenAI"
Evaluation details
Result
Evaluator
Details
Meta Data
100.00%
Matches word count
n/a
neededClean
false
words
500
49.37%
Dialogue to Total Word Ratio
Ratio: 26.24%, Deviation: 3.76%
neededClean
false
wordsTotal
503
wordsDialogue
132
74.6826%