NC Bench
Overview
About
Tests
Models
Vendors
Run: 2026-02-16-01KHM60VYG76BMX7AW7494WA6F
Run Details
Model:
DeepSeek V3.2
Test:
Dialogue tags
Scenario:
Write 500 words with 70% dialogue
Duration:
34.5s
Cost:
$0.000342
Provider:
Novita
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
The air in the tiny, cluttered safehouse smelled of stale coffee and desperation. Bill was frowning at a hand-drawn map spread on the table, a child’s crayon rendering of a suburban backyard. “Okay, so the target location is a ‘Fairy Wonderland,’” he said, tapping the paper. “Primary ingress is the back gate, here, which will be covered in crepe paper streamers. Our point of entry is the garden shed, which will be doubling as a ‘Gnome Depot’ for the afternoon.” Molly peered over his shoulder, twirling a lockpick absentmindedly. “Security?” “Tight. Two adult supervisors, likely caffeine-deprived and highly volatile. A rotating patrol of sugar-fueled ankle-biters, ages four to seven. Our biggest threat is a miniature schnauzer named Mr. Snuffles, who reportedly naps from two to four, but we can’t count on that.” “Weapons?” “We’re going non-lethal. I’ve got a supersoaker filled with liquid melatonin. You’ll have the bubble wand laced with a mild, tear-free soap solution. Creates a diversion and impairs vision.” Molly nodded, all business. “The loot?” Bill produced a photograph. It showed a wobbly tower of brightly wrapped boxes. “Primary objective: the big one with the rocket ship paper. Intel suggests it’s a limited-edition, remote-control X-7 Falcon. High resale value. Secondary targets: anything rectangular and heavy. Avoid the soft, squishy parcels. Plush toys are a logistical nightmare; low profit, high volume.” “Extraction route?” “Through the ‘Enchanted Forest,’ which is Mrs. Henderson’s prize-winning hydrangea bushes. We cross the vegetable patch—watch for tripwires, I mean, cucumber vines—and over the low fence by the compost bins. The getaway vehicle, a nondescript minivan, will be idling two blocks over, masquerading as a soccer mom’s transport.” Molly chewed her lip. “What about the birthday boy? The… six-year-old ‘mark’?” “Codenamed ‘The Tyrant.’ He’ll be the epicenter of the chaos, blinded by the cake-lighting ceremony. We hit during the singing. The auditory cover will be perfect. All adults will be filming, backs turned. It’s a three-minute window.” “And if we’re made?” “Contingency Plan Alpha: You pretend to be a lost clown. I’ll start enthusiastically assembling lawn furniture. They’ll just think we’re weird relatives.” Molly finally cracked a smile. “I don’t know, Bill. This feels… low. Even for us. Stealing from a kid?” Bill sighed, his shoulders slumping. “You think I like this? My therapist is going to have a field day. But the McAllister job fell through, the diamond heist was a fiasco because you’re allergic to guard dogs, and we owe Big Tony a lot of money. This is a soft target with a high yield of essentially untraceable, cash-convertible goods.” He pointed a stern finger at the photo of the rocket-ship present. “That’s not a toy, Molly. That’s our vig. That’s not a six-year-old’s birthday; it’s a poorly guarded distribution center. Now, do you want to explain to Big Tony why we’re using his money to fund a fairy wonderland?” Molly stared at the crayon map, at the little drawing of a cake. She squared her shoulders. “Fine. But I’m taking a slice of cake on the way out. Operational calories.” “Deal,” Bill said, rolling up the map. “We move at fourteen-hundred hours. Happy birthday to us.”
created_at
"2026-02-16T21:35:11.282943+00:00"
2/16/2026, 9:35:11 PM
model
"deepseek/deepseek-v3.2-20251201"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
1211
moderation_latency
(null)
generation_time
24363
tokens_prompt
77
tokens_completion
907
native_tokens_prompt
67
native_tokens_completion
811
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
0.000342423
router
(null)
provider_responses
0
status
(null)
is_byok
false
latency
10077
endpoint_id
"b39a588a-b338-43ba-b046-226b9e24afa8"
provider_name
"SiliconFlow"
model_permaslug
"deepseek/deepseek-v3.2-20251201"
1
id
"a64015045b2edf656de1d17b7cb3e2d1"
status
200
is_byok
false
latency
1211
endpoint_id
"8a5fb6d3-ef71-4958-af54-269ffa3a3c6e"
provider_name
"Novita"
model_permaslug
"deepseek/deepseek-v3.2-20251201"
api_type
"completions"
id
"gen-1771277676-1eAwp1Fh5nHnN6pwQvJM"
upstream_id
"a64015045b2edf656de1d17b7cb3e2d1"
total_cost
0.000342423
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Novita"
Evaluation details
Result
Evaluator
Details
Meta Data
0.0000%
Matches word count
n/a
neededClean
false
words
544
1.6743%
Dialogue to Total Word Ratio
Ratio: 79.04%, Deviation: 9.04%
neededClean
false
wordsTotal
544
wordsDialogue
430
0.8371%