NC Bench
Overview
About
Tests
Models
Vendors
Run: 2026-02-16-01KHM5K9HJA48E94R2E7Y5K9KX
Run Details
Model:
DeepSeek V3 (2025-03-24)
Test:
Dialogue tags
Scenario:
Write 500 words with 70% dialogue
Duration:
14.9s
Cost:
$0.000558
Provider:
AtlasCloud
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 leaned over the diner table, stirring her lukewarm coffee with a plastic straw. “Listen, Bill, I’ve got the perfect score. A low-risk, high-reward situation. Security’s lax, entry’s easy, and the haul? Untraceable.” Bill squinted at her, his bacon grease-stained fingers tapping the table. “You’re talking about the toy store again, aren’t you? I told you, the laser tag system’s too—” “No, you big dumb oaf,” Molly hissed, glancing around before lowering her voice. “The Benson twins’ birthday party. Saturday. Their parents rented out the whole community center.” Bill blinked. “You wanna rob a kid’s party?” “Not the party. The presents, Bill. Piles of ‘em. iPads, Lego sets, those creepy robot dogs that bark—no serial numbers, no tracking. Pure profit.” Bill scratched his stubble. “So we’re what? Dressing up as clowns?” “Clowns are creepy. No, we go in as caterers.” “We don’t know how to cater.” “Neither do most caterers,” Molly shot back. “You show up in a hairnet, and people just assume you’re supposed to be there. We schmooze, blend in, then load the gifts into a ‘supply van’ right out the back.” Bill frowned. “What about the kids? They’ll notice their presents vanishing.” “You’ve clearly never been to a kid’s party,” Molly smirked. “They’re hopped up on sugar and chaos. By the time they realize, we’re halfway to the buyer.” Bill leaned back, considering. “What’s the exit strategy if someone catches on?” Molly waved a dismissive hand. “Oh please. Worst case, we fake a food allergy. You start wheezing, I scream about gluten—every parent within five miles will panic and forget we exist.” A slow grin spread across Bill’s face. “Okay. But one condition.” “Name it.” “I get first pick of the loot.” Molly narrowed her eyes. “You just want that stupid drone from the TV ads, don’t you?” Bill’s cheeks flushed. “It’s got night vision!” Molly groaned. “Fine. But I’m keeping the robot dog.” “Why?” “Because it’s creepy,” Molly grinned. “Perfect for distracting the kids while we work.” Bill raised his coffee cup. “To the easiest heist in history.” Molly clinked hers against it. “And the most ridiculous.” As they plotted the finer details—napkins as blueprints, ketchup packets marking guard rotations—neither noticed the waitress refilling their cups, her eyebrow arched just slightly higher than before.
created_at
"2026-02-16T21:27:27.426767+00:00"
2/16/2026, 9:27:27 PM
model
"deepseek/deepseek-chat-v3-0324"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
935
moderation_latency
(null)
generation_time
14892
tokens_prompt
77
tokens_completion
699
native_tokens_prompt
67
native_tokens_completion
618
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
0.000557664
router
(null)
provider_responses
0
id
"a39a2096d49b44f89cdb88f8930b5b66"
status
200
is_byok
false
latency
935
endpoint_id
"5bb598f5-fd97-4f8b-91e4-ceab3743a03b"
provider_name
"AtlasCloud"
model_permaslug
"deepseek/deepseek-chat-v3-0324"
api_type
"completions"
id
"gen-1771277231-czPKzuKPiRtLRWd2Ex2r"
upstream_id
"a39a2096d49b44f89cdb88f8930b5b66"
total_cost
0.000557664
cache_discount
6.48e-7
upstream_inference_cost
0
provider_name
"AtlasCloud"
Evaluation details
Result
Evaluator
Details
Meta Data
0.0000%
Matches word count
n/a
neededClean
false
words
385
97.9138%
Dialogue to Total Word Ratio
Ratio: 70.65%, Deviation: 0.65%
neededClean
false
wordsTotal
385
wordsDialogue
272
48.9569%