NC Bench
Overview
About
Tests
Models
Vendors
Run: 2026-02-16-01KHM5YSH9Y3SAJMD6TMZ52990
Run Details
Model:
DeepSeek V3.2
Test:
Dialogue tags
Scenario:
Write 500 words with 50% dialogue
Duration:
20.3s
Cost:
$0.000329
Provider:
Novita
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
The warehouse loft smelled of dust and desperation. Molly traced a blue-inked finger over the schematics spread on the rickety table. “It’s a fortress, Bill. A pastel-colored, balloon-anchored fortress.” Bill peered over her shoulder, squinting at the layout of the suburban McMansion. “Fortress? It’s a kid’s birthday party. The biggest threat is a sugar-crazed seven-year-old with a plastic sword.” “Precisely! Unpredictable, miniature chaos agents. Our intel says the theme is ‘Jungle Explorers.’ That means camouflage netting, artificial foliage, and a hired guy in a mangy gorilla suit. Perfect visual cover.” “And the extraction point?” Bill asked, tapping the diagram of the backyard. “Slip through the side gate during the piñata frenzy. Every adult eye will be on a suspended papier-mâché donkey. We blend, we grab, we go. The presents will be piled in the sunroom, here.” Her finger landed on a glass-walled room off the kitchen. “Blend in? Molly, we’re in our forties. I have a tattoo of a dragon eating a motorbike.” “So wear a safari hat. Look like a disgruntled uncle. I’ll be the quirky aunt no one talks to. We’ll case the gifts during the ‘Happy Birthday’ song. Emotional peak, zero vigilance.” Bill paced, his boots scuffing the concrete. “The payload. What’s the take?” Molly pulled a crumpled list from her pocket. “Estimated value based on registry scans: a next-gen gaming console, three sets of noise-cancelling headphones—parental gold—a junior chemistry set with legitimate, mildly explosive potential, and a staggering amount of branded plastic. The real score is the ‘Limited Edition Star-Counting, Robo-Hamster.’ It’s reselling for six hundred bucks online.” “A Robo-Hamster.” Bill deadpanned. “We’ve sunk to stealing robot rodents.” “It’s a metaphor for the emptiness of consumer culture, Bill. Now focus. Exfiltration is the weak point. We can’t just waltz out with a sack of toys.” “Cake delivery,” Bill said, a slow grin spreading. “We go in empty-handed, we leave with a giant, empty cake box. Stack the loot inside, walk out like catering staff. Who questions the people with the empty cake box? It’s a symbol of mission accomplished.” Molly considered this, a flicker of respect in her eyes. “Not bad. But the box will be sticky.” “All great heists have a sticky element. What about the parents? The Johnsons.” “Karen Johnson runs a tight ship, but Mike Johnson is the variable. He’ll be manning the grill, drinking craft beer, and trying to forget his mortgage. He’s our distraction. I’ll ask him to explain the difference between ales and lagers. That’ll give us a solid twelve-minute window of him pontificating.” Bill nodded, finally looking convinced. “Alright. But I have one non-negotiable condition.” “Name it.” “We take a piece of the cake. A big corner piece with the fondant jungle explorer hat on it. A professional job deserves a professional perk.” Molly smiled for the first time. “Deal. We hit during the gorilla’s conga line. Chaos is a ladder, Bill.” “To a pile of overpriced toys,” he sighed, rolling up the plans. “Let’s go shopping for a safari hat. Do they make them in extra-cynical?”
created_at
"2026-02-16T21:33:49.106073+00:00"
2/16/2026, 9:33:49 PM
model
"deepseek/deepseek-v3.2-20251201"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
1024
moderation_latency
(null)
generation_time
20306
tokens_prompt
77
tokens_completion
875
native_tokens_prompt
67
native_tokens_completion
777
native_tokens_completion_images
(null)
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
"stop"
usage
0.000328823
router
(null)
provider_responses
0
id
"ed9731650bb4ee88b2bc3e4090cc3943"
status
200
is_byok
false
latency
1024
endpoint_id
"8a5fb6d3-ef71-4958-af54-269ffa3a3c6e"
provider_name
"Novita"
model_permaslug
"deepseek/deepseek-v3.2-20251201"
api_type
"completions"
id
"gen-1771277608-E7smEeMWTVAadhPavpl8"
upstream_id
"ed9731650bb4ee88b2bc3e4090cc3943"
total_cost
0.000328823
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Novita"
Evaluation details
Result
Evaluator
Details
Meta Data
1.0360%
Matches word count
n/a
neededClean
false
words
526
0.0000%
Dialogue to Total Word Ratio
Ratio: 80.23%, Deviation: 30.23%
neededClean
false
wordsTotal
526
wordsDialogue
422
0.5180%