NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-18-01KM08TMGMEEKEJZQ79MNJPP1G
Run Details
Model:
GPT-5.4 Mini (Reasoning, Low)
Test:
Dialogue tags
Scenario:
Write 500 words with 70% dialogue
Duration:
6.0s
Cost:
$0.004383
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
7.03%
Dialogue to Total Word Ratio
(details)
3.5171%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 70% 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
“Bill, you’re doing that thing with your eyebrows again,” Molly whispered, flattening herself against the bakery wall across from the community hall. “What thing?” “The thing where you look like a collapsed accordion.” “I am thinking,” Bill said, peering through the rain-streaked window at the glittering chaos inside. “A professional thief needs a face that says ‘I belong here’ and not ‘I just lost a fight with a lamp.’” “You do belong here,” Molly said. “You’re on the list.” “I’m on the list as ‘Uncle Barry.’” “An inspired disguise.” “I borrowed the name from your ex.” “My ex was named Barry?” “You said your ex was a liar.” “That narrows it down to the entire human population.” Bill adjusted the cheap paper mustache he’d glued above his lip. It was already peeling at the corners. “How do I look?” “Like a man whose life choices were made by an angry raccoon.” “That bad?” “Worse. But in a charming way.” Bill squinted at the hall. “There are balloons. Why are there always balloons?” “Because children think joy should be airborne,” Molly said. “Focus. The cake is there, the gifts are there, and so are the parents, which is unfortunate because parents have the sort of vigilance usually reserved for prison guards and hawks.” Bill nodded. “And our goal is to steal all the presents.” “All the presents.” “From a children’s party.” “Yes.” “In broad daylight.” “It’s after lunch.” “That’s not better.” Molly smiled without warmth. “Listen, Bill, this is a perfect opportunity. Everyone is distracted by party games, tiny sandwiches, and the kind of music that sounds like it was written by a sugar-addled alien.” “You said it was a perfect opportunity to steal from a bank last month.” “It was.” “And then it turned out to be a pet grooming salon.” “They had a vault vibe.” “They had shampoo.” Molly checked her watch. “No nostalgia. We go in with the crate of fake party favors, we swap the labels, and we leave with every wrapped box, every bag, every suspiciously lumpy gift sack in sight.” Bill frowned. “What about the one from Aunt Linda?” “What about it?” “She knitted something.” “Even better. Less noise.” “She may have knitted a sweater.” “We can pawn a sweater.” “She may have knitted eleven sweaters.” “Then we can open a knitwear boutique.” Bill stared at her. “You plan to steal presents from children and start a business?” “Bill, that’s called diversification.” A burst of laughter spilled from the hall, followed by the tinny squeal of a clown horn. Bill flinched. “I hate clowns.” “You hate anyone with confidence.” “I distrust confidence in face paint.” Molly leaned in. “Are you in or not?” Bill took a breath, straightened his rented blazer, and said, “I’m in, but I want to note for the record that this is morally complicated.” “Noted.” “And also structurally absurd.” “Noted.” “And if one child cries, I will feel awful.” “Also noted. If one child cries, we leave extra quickly.” Bill blinked. “That’s your line?” “It’s compassion with pacing.” He looked back through the window. “There’s a boy with a pirate hat opening a present. It’s a drum.” “Perfect.” “How is that perfect?” “Because if we’re caught, we can blame the drum.” Bill stared at her. “You are a terrible person.” “I know. That’s why you hired me.” “I did not hire you.” “You proposed the idea with your eyes.” “My eyes are very expressive.” “They’re saying, ‘Please let me do crime, but elegantly.’” Bill sighed, picked up the tray of fake cupcakes, and grinned. “Fine. But if we get away with this, I’m naming our next job after a musical instrument.” “Why?” “Because of the drum.” Molly patted his shoulder. “Bill, if this works, we’ll be legends.” “And if it doesn’t?” She glanced toward the party, where a mother in sequins was turning with the alertness of a border collie. “Then, my dear accomplice, we will become very educational examples.” Bill swallowed. “Right. Legends or lessons.” “Exactly.” He pushed the door open. “Let’s go rob a birthday.”
created_at
"2026-03-18T10:45:54.848Z"
3/18/2026, 10:45:54 AM
model
"openai/gpt-5.4-mini-20260317"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
2317
moderation_latency
(null)
generation_time
5783
tokens_prompt
77
tokens_completion
1360
native_tokens_prompt
66
native_tokens_completion
963
native_tokens_completion_images
0
native_tokens_reasoning
35
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
0.004383
router
(null)
provider_responses
0
endpoint_id
"9ee065b2-3d1c-43bc-bdd7-28af3b148282"
id
"resp_0d064fab1d7710810169ba8262e648819699295a555f23eb15"
is_byok
false
latency
181
model_permaslug
"openai/gpt-5.4-mini-20260317"
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-1773830754-GEViDMwOIon2PbTAzhnK"
upstream_id
"resp_0d064fab1d7710810169ba8262e648819699295a555f23eb15"
total_cost
0.004383
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenAI"
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Matches word count
n/a
neededClean
false
words
676
7.03%
Dialogue to Total Word Ratio
Ratio: 77.29%, Deviation: 7.29%
neededClean
false
wordsTotal
678
wordsDialogue
524
3.5171%