NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-18-01KM0A4WBS93BA2S2SDAFNBWD4
Run Details
Model:
GPT-5.4 Nano
Test:
Dialogue tags
Scenario:
Write 500 words with 50% dialogue
Duration:
5.4s
Cost:
$0.001439
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
37.93%
Dialogue to Total Word Ratio
(details)
18.9636%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 50% 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 adjusted her ski mask in the car’s rearview like it was a stage costume. “Bill, I told you—soft shoes. Quiet. Like a stealthy cloud.” Bill held up a pair of paw-shaped slippers. “These are not clouds, these are woodland predators.” Molly whispered, “Predators get caught. We are… seasonal burglars. Friendly. Festive.” Bill nodded solemnly. “Yes. We steal Christmases. With manners.” Molly glanced at the address on a crumpled map. “The Smithhouse. It’s a children’s party. There will be juice boxes, balloons, and—” “Presents,” Bill finished, grinning. “A veritable treasure chest of tiny socks and suspiciously expensive electronics.” Molly leaned closer. “We’re not here for electronics.” Bill blinked. “Why did you pick up the list, then? Because I like the list.” “It’s for… sentimental reasons,” Molly said. “Also, we will take something practical. Like the gift cards.” Bill clapped once, then immediately clapped his own mouth. “Gift cards. The most portable joy.” Molly checked her watch. “Okay. Plan. We enter through the side gate, cut the motion sensor string—” Bill interrupted, “With the little scissors we definitely didn’t steal from the craft store.” Molly’s eyes narrowed. “Those scissors were on sale.” Bill shrugged. “So were we. Like… morally.” Molly reached into her bag and produced a roll of duct tape. “We bypass the laser line using this.” Bill pointed. “That duct tape says ‘heavy duty’.” Molly smiled. “That’s for thieves.” Bill’s smile widened. “I thought it was for boats.” “It’s for anyone with determination,” Molly said. “Now, once inside, we head straight to the living room where the gifts are stacked.” Bill traced an invisible line in the air. “Living room, then loot, then—” Molly raised a finger. “No shouting. No cackling. No touching anything that smells like play-dough.” Bill nodded. “I can cackle silently. It’s a skill.” Molly sighed. “Try it outside, please.” Bill lifted a hand. “I brought a decoy.” Molly stared. “You brought a decoy to a children’s party?” “It’s small,” Bill said, holding up a plush dinosaur. “It looks harmless.” Molly tilted her head. “Dinosaur plush don’t scare me.” Bill whispered, “The real decoy is me wearing this.” He pulled down a reflective vest that read IMPORTANT DELIVERY. It also had stickers of cartoon stars. Molly stared at it for a full second. “Bill. You look like a substitute teacher in a heist movie.” Bill beamed. “That’s the goal. Trust the vibe.” Molly exhaled. “Trust the vibe gets you detained.” Bill leaned in. “Detained? No. Appreciated. By the parents. Like, ‘Wow, the helpful delivery person!’” Molly tapped the map. “We are helpful burglars. We are not criminals.” Bill blinked. “We’re stealing presents.” Molly’s tone softened. “We’re… borrowing holiday magic until the rightful owners receive replacements.” Bill stared. “Are we replacing them?” Molly hesitated, then said, “We’re replacing them with other presents we already had.” Bill’s face fell. “So we’re doing re-gifting.” Molly nodded briskly. “Re-gifting is legal. It’s basically charity with receipts.” Bill considered this. “Charity with receipts is my love language.” A car alarm across the street chirped once, like a nervous bird. Both thieves froze. Bill whispered, “Did it hear us?” Molly whispered back, “No, it’s just… auditioning.” Bill swallowed. “Okay. Time. We move when the porch light clicks.” Molly checked the house. The porch light blinked, and in the flicker, the yard looked like a cheerful trap made of streamers. Bill pointed at the side gate. “After you.” Molly stepped forward. “No, after you. I’m the one with the tape.” Bill stepped too, then paused. “Molly?” “What,” she hissed, already crouching. Bill gestured at the air. “I’m hearing jingle bells.” Molly listened. Faint chimes floated from inside the house, tangled with laughter. “Those are probably… toys,” Molly said. Bill looked genuinely alarmed. “Toys that jingle are never innocent.” Molly tugged the duct tape out with a practiced rip. “All right. Quiet. Friendly. Fetching presents.” Bill grinned again. “Like Santa’s extremely underpaid cousins.” Molly crawled toward the motion sensor, voice barely there. “On three. One—” Bill whispered, “Two—” Molly: “Three.” They slipped into the side yard together, moving like embarrassed ninjas at a birthday party, and the house greeted them with one more chirp, as if the lights themselves were asking, “So… where’s the loot?”
created_at
"2026-03-18T11:08:59.14Z"
3/18/2026, 11:08:59 AM
model
"openai/gpt-5.4-nano-20260317"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
470
moderation_latency
(null)
generation_time
5115
tokens_prompt
77
tokens_completion
1304
native_tokens_prompt
66
native_tokens_completion
1141
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
0.00143945
router
(null)
provider_responses
0
endpoint_id
"0c835f2e-c18d-4e8c-b245-e1e3bd08b97f"
id
"resp_0d7d438a2a4ebaad0169ba87cb37308194931419f0110cb991"
is_byok
false
latency
207
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-1773832139-dhnm3QLPTIDqy9IiIDKt"
upstream_id
"resp_0d7d438a2a4ebaad0169ba87cb37308194931419f0110cb991"
total_cost
0.00143945
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
702
37.93%
Dialogue to Total Word Ratio
Ratio: 54.40%, Deviation: 4.40%
neededClean
false
wordsTotal
704
wordsDialogue
383
18.9636%