NC Bench
Overview
About
Tests
Models
Vendors
Run: 2026-02-16-01KHK964WWPK63NYYCCZ0Z3QJ7
Run Details
Model:
GPT-5
Test:
Dialogue tags
Scenario:
Write 500 words with 50% dialogue
Duration:
1m 27s
Cost:
$0.064783
Provider:
OpenAI
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
Behind the community center, wedged between a bouncy castle and a dumpster perfumed with frosting, Molly reviewed a crumpled blueprint while Bill practiced quiet jazz hands with very loud feet. "So we are robbing a birthday party, and you wore tap shoes?" Molly pinched the bridge of her nose, staring at his shoes like they owed her rent. "Distraction, obviously; the clatter says confidence while my pockets whisper opportunity." Balloons bobbed above the back door; a handmade sign warned no adults without children, which felt personally insulting. "Confidence is great, but toddlers smell fear and frosting, not existential bravado." Bill, unfazed, dabbed vanilla on his wrists like cologne and grinned at their reflection in the window, two misfits rehearsing authority. "Then we smell like frosting; I brought vanilla extract and questionable wisdom." Molly tucked a roll of stickers into her sleeve, the top one reading Recall in festive script, and snapped a rubber band around her clipboard. "Focus. The presents are stacked by the magician, who looks suspiciously competent." Music thumped from the hall, a remix of nursery rhymes that felt like being chased by alphabet soup. "Competent magicians store doves in sleeves; they can store decoys for us." The magician’s top hat sat on a crate, feathers peeking out, while a suspiciously large rabbit lounged like union labor. "We need a cover. Balloons, a clipboard, and the brisk shambling of authority." Bill stamped once; the echo rattled a row of juice boxes, which hissed like a tiny audience. "I can shambling; authority might squeak under these shoes, but I am committed." She flashed their forged badges: laminated breakfast cereal coupons, tastefully trimmed into bureaucratic rectangles. "We arrive as Gift Inspectors, Department of Ribbons, citing Code Twelve: Excessive Glitter." Somewhere inside, a chorus counted down to candles; the lights would drop, the cameras would rise, and chaos would bloom like confetti. "Perfect. I carry the clipboard; you carry the gravitas and the giant sack." Molly hefted the burlap sack, rechristened repository with a marker, and practiced her stern nod in the glass. "We do not say giant sack; we say repository for unallocated merriment." Bill mimed stamping forms, his elbows articulating like a marionette on cross-examination. "Right, repository. We confiscate anything that jingles, squeaks, or threatens structural integrity." He angled his hat to Responsible and stuffed a pen behind an ear, for symmetry and swagger. "Timing matters. During cake, chaos peaks, and parents photograph each other pretending calm." "I will pose near the candles, generating smoke cover with sincere applause." "We tag the gifts with stickers, red for recall, blue for bureaucratic destiny." "And yellow for banana shaped hazards, which I personally vow to investigate vigorously." "No snacks. We are professionals, and crumbs are the snitches of carpeting." "Counterpoint: cupcakes are camouflage; frosting mustache equals trusted uncle with clipboard." "Trusted uncles do not wear tap shoes that scream jazz hands at every step." "Fine, I will tiptoe like a remorseful cymbal; listen for silent thunder." They listened. "Absolutely no improvisation."
created_at
"2026-02-16T13:12:08.899974+00:00"
2/16/2026, 1:12:08 PM
model
"openai/gpt-5-2025-08-07"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
3232
moderation_latency
(null)
generation_time
86732
tokens_prompt
77
tokens_completion
1904
native_tokens_prompt
66
native_tokens_completion
6470
native_tokens_completion_images
0
native_tokens_reasoning
5760
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
0.0647825
router
(null)
provider_responses
0
id
"resp_0ab5dff3e877cd0f0169931750f488819392d119dbc60a7e54"
status
200
is_byok
false
latency
238
endpoint_id
"7c2f859a-7890-4e8e-b1de-1cd1c0a800b4"
provider_name
"OpenAI"
model_permaslug
"openai/gpt-5-2025-08-07"
api_type
"completions"
id
"gen-1771247440-ZxOjyebixOoE8AHVfy7G"
upstream_id
"resp_0ab5dff3e877cd0f0169931750f488819392d119dbc60a7e54"
total_cost
0.0647825
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenAI"
Evaluation details
Result
Evaluator
Details
Meta Data
99.9990%
Matches word count
n/a
neededClean
false
words
501
99.9502%
Dialogue to Total Word Ratio
Ratio: 49.90%, Deviation: 0.10%
neededClean
false
wordsTotal
501
wordsDialogue
250
99.9746%