NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-18-01KM09Z3KD9FVA3JCJ70BXQZ01
Run Details
Model:
GPT-5.4 Nano
Test:
Dialogue tags
Scenario:
Write 500 words with 70% dialogue
Duration:
8.0s
Cost:
$0.001917
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
5.95%
Dialogue to Total Word Ratio
(details)
2.9753%
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
“Molly, are you sure this is the right house?” Bill whispered, crouched beside the hedges like a particularly nervous lawn ornament. “Because if I die from a landscaping-related injury, I’m haunting you professionally.” “Relax, Bill,” Molly said, sliding a thin strip of metal into her lockpick case. “This is the place. I’ve seen the address on the invitation. Also, the inflatable dinosaur in the yard gave it away.” “The inflatable dinosaur gave it away,” Bill repeated. “Great. Nothing says ‘subtle crime’ like a foam T. rex.” “Shh,” Molly hissed, listening. “Do you hear that?” “Laughter?” Bill asked. “A clown? A kazoo orchestra? The sound of capitalism being shredded?” Molly smiled. “Exactly. Kids are inside. Parents are distracted. It’s like robbing a bank, but with juice boxes.” Bill leaned closer. “We’re stealing presents from children.” “Technically,” Molly corrected, “we’re reallocating holiday wealth from greedy gift-goblins to deserving criminals.” Bill blinked. “Greedy gift-goblins.” “Mhm.” Molly popped the lock with a quiet click. “Now, remember the plan. We go in, we grab everything, we leave no mess, and we do it with style.” “Style,” Bill said. “I brought the style. I have the silent shoes.” “You do, actually,” Molly admitted, impressed. “But those shoes sound like two angry mice. Step lighter.” Bill lifted his foot and set it down again, slowly. “Like this?” “Like you’re trying to sneak up on a cake,” Molly said. “I can sneak up on cake.” Bill grinned. “I once followed a donut trail for three blocks.” Molly pushed the garden gate open. “Move. The frosting economy awaits.” They slipped across the yard. Inside, music thumped faintly—something cheerful with too many claps. Bill pointed through the window. “Okay. I see the tree. I see the pile of presents. I see the… glowing unicorn table?” “Don’t stare,” Molly murmured. “Staring makes you a participant in destiny.” Bill cranched his neck. “I’m not staring. I’m… surveying. Like a military man.” “Military men don’t steal toys from toddlers,” Molly said. “Militaries absolutely steal toys from toddlers,” Bill argued. “War is just organized improvisation.” Molly shot him a look. “Fine. Just don’t improvise near the glitter.” “What if we improvise far away from the glitter?” Bill asked. “Bill.” “What?” He lowered his voice to a whisper. “I’m asking questions to reduce risk.” Molly adjusted her backpack. “Our goal is simple: wrap the gifts. Load the bags. Escape before the ‘party game’ starts. You know the one where they freeze and then scream when the music stops.” Bill nodded solemnly. “Musical statues of doom. I hate it. I once froze so hard my soul left my body.” “Then don’t freeze,” Molly said. “Just move like you belong there.” “I belong there emotionally,” Bill said, already holding a small flashlight like it might tell him a comforting secret. “I have the heart of a child. Specifically, a child who loves stealing.” Molly pushed the window latch up. “We’re going in through here. Quietly. Like librarians doing a heist.” “Librarians are very scary,” Bill said. “They have rules. I hate rules.” “Exactly,” Molly said. “Now shut up and follow.” Bill squeezed through the window, muttering, “Okay, okay, okay,” like the walls were a set of angry bouncers. “I’m inside. I’m inside. Is it normal for my knees to complain this loudly?” “Stop complaining,” Molly said, and her voice softened. “We’re almost done. Then we’ll be legends.” Bill’s whisper turned excited. “Legends who stole presents.” “Legends who restored balance,” Molly corrected again. “And also legends who never get caught.” Bill pointed to a room filled with bright wrapping paper and balloons shaped like numbers. “I see the ‘From Santa’ corner.” Molly’s eyes widened. “Perfect. That’s the real haul.” Bill swallowed. “Santa’s gifts.” “Santa’s insurance policy is probably huge,” Molly said. “Let’s do it.” Bill crept toward the pile like a professional nap thief approaching a sleeping baby. “Molly, if anyone asks, we’re… delivery.” “Delivery,” Molly agreed. “Yes. We’re delivery. Loud, clumsy delivery.” Bill glanced at her. “That doesn’t sound quiet.” “It sounds like chaos,” Molly said. “Chaos is quiet if you’re fast.” Bill picked up a package wrapped in sparkly paper. “This one’s heavy. Probably a dinosaur set. Or a book about dinosaurs. Or… both.” “Grab everything with pictures,” Molly instructed. “Kids love pictures. They also love repeating themselves. It’s adorable, but also a distraction.” Bill laughed under his breath. “You’re going to make me sentimental.” “Don’t get sentimental,” Molly said. “Get efficient. I’ll keep watch.” Bill held up his flashlight. “You keep watch by watching, or by doing that thing where you stare intensely at nothing like a haunted office plant?” Molly frowned. “I do not do that.” Bill leaned back, listening. “Yes, you do. It’s unnerving. Like… like you’re monitoring the shadows for suspicious thoughts.” Molly hissed, “Less talk. More steal.” Bill nodded quickly. “Yes, ma’am. Steal. With gusto. I’m stealing with gusto.” Molly took another step, then paused. “Bill.” “Yeah?” Bill whispered. “Did you just step on a squeaky toy?” Bill froze. “No.” A squeak echoed anyway, loud enough to make a distant parent laugh. Molly stared. “Bill.” Bill swallowed. “That wasn’t me.” The music in the room abruptly changed, and a child’s voice sang, “Ready… set… freeze!” Molly’s eyes widened. “Okay. New plan.” Bill gulped. “New plan?” “We freeze,” Molly said, dead serious. Bill blinked. “We’re criminals. We freeze.” Molly nodded. “Just long enough for the game to pass.” Bill tried to hold still. “I can freeze.” The squeaky toy squeaked again. Molly leaned in, teeth clenched. “Bill.” Bill whispered, barely moving. “The toy is freezing. I’m not.” Molly looked ready to burst into laughter and fury simultaneously. “If we get caught, we’re not explaining ourselves.” Bill managed a tiny grin. “If we get caught, we’re blaming Santa.” “Fine,” Molly whispered. “But hurry. On three, we run.” Bill’s eyes went wide. “We run on three.” Molly counted under her breath. “One… two…” Bill whispered, “What if the unicorn table chases us?” Molly hissed, “Three!” They darted forward in a blur of silent shoes and wrapped chaos, grabbing gifts like they’d been training their whole lives for this exact ridiculous moment.
created_at
"2026-03-18T11:05:49.946Z"
3/18/2026, 11:05:49 AM
model
"openai/gpt-5.4-nano-20260317"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
1296
moderation_latency
(null)
generation_time
7577
tokens_prompt
77
tokens_completion
1918
native_tokens_prompt
66
native_tokens_completion
1523
native_tokens_completion_images
0
native_tokens_reasoning
0
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
0.00191695
router
(null)
provider_responses
0
endpoint_id
"0c835f2e-c18d-4e8c-b245-e1e3bd08b97f"
id
"resp_061feb1fb4bdec7a0169ba870e55d48190a1c2bb21f0861ac0"
is_byok
false
latency
259
model_permaslug
"openai/gpt-5.4-nano-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-1773831949-89X6S4pKISfPkKbHF6cL"
upstream_id
"resp_061feb1fb4bdec7a0169ba870e55d48190a1c2bb21f0861ac0"
total_cost
0.00191695
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
1018
5.95%
Dialogue to Total Word Ratio
Ratio: 62.49%, Deviation: 7.51%
neededClean
false
wordsTotal
1021
wordsDialogue
638
2.9753%