NC Bench
Overview
About
Tests
Models
Vendors
Run: 2026-02-16-01KHM5SS8A533G6CN19HVMJ5P1
Run Details
Model:
DeepSeek V3.1
Test:
Dialogue tags
Scenario:
Write 500 words with 70% dialogue
Duration:
1m 45s
Cost:
$0.000735
Provider:
DeepInfra
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 back room of the diner smelled of stale grease and desperation. Molly stirred her coffee with a grimace, watching the thin brown liquid swirl. Across the Formica table, Bill was meticulously sketching on a napkin with a stolen crayon. “It’s a hard target, Moll. A real nightmare,” Bill said, not looking up from his crude diagram. “Perimeter is soft, but the interior is a logistical quagmire. High foot traffic, unpredictable patrols, and the primary assets are all under direct surveillance.” Molly sighed, pushing her cup away. “Bill, for the last time, it’s a six-year-old’s birthday party at the community hall. The ‘perimeter’ is a crepe paper streamer. The ‘unpredictable patrols’ are a sugar-crazed kid named Timmy on a Big Wheel.” “See? You’re already underestimating him!” Bill jabbed the crayon at her. “That’s how they get you. Little Timmy is the wild card. He’s the loose cannon, a pint-sized liability on three wheels. Our exit strategy is compromised before we even breach.” “We ‘breach’ by walking in the front door wearing party hats,” Molly said flatly. “We’re ‘Uncle Bill’ and ‘Aunt Molly,’ here for little… what’s the brat’s name again?” “Cayden,” Bill whispered, as if saying a forbidden word. “Or maybe Kayden. The intel is fuzzy. Point is, the presents will be piled near the cake table. That’s the motherlode. But the approach is wide open. We need a diversion.” “A diversion. At a child’s party.” “Exactly.” Bill’s eyes gleamed. “I’ve been workshopping a few ideas. Option one: I fake an allergic reaction to the piñata. Convincing, theatrical, lots of screaming.” “There’s no peanut piñata, Bill.” “Option two: we release a crate of crickets. Chaos. Panic. While they’re herding orthopterans, we’re herding gift bags.” Molly leaned forward, her voice low and serious. “Or, and hear me out, I will create a diversion by asking the magician where he bought his sequined waistcoat. You will then casually stack the presents into the large, empty gift bag we brought specifically for this purpose. We say our goodbyes, thank ‘Kayden’ for the lovely time, and we leave. No crickets. No medical theatrics.” Bill looked crestfallen. “It’s… it’s a simple smash-and-grab. Where’s the artistry? The finesse?” “The ‘artistry’ is not getting caught and having to explain to a room full of horrified parents why we’re stealing a Lego Duplo set and a dozen gift cards to Build-A-Bear.” She reached across and took his hand, the one not holding the crayon. “This isn’t the Frankfurt Diamond Exchange. It’s a kid’s party. The biggest risk is getting glitter on your jacket or being forced to sing the Happy Birthday song.” He stared at his napkin blueprint, a masterpiece of misguided ambition. “The Happy Birthday song,” he muttered. “The ultimate test of a thief’s nerve. Your pitch is always flat, you know. It’s a tell.” “I’ll work on my pitch. You work on your ‘jolly uncle’ smile. It currently looks like you’re sizing up the room for valuables.” “It’s my concentrating face!” “It’s the same face you made before we tried to knock over that armored car.” Molly stood up, dropping a few bucks on the table. “The plan is simple. We go in, we smile, we grab the loot, we leave. Understood?” Bill crumpled the napkin into a ball with a sigh of profound resignation. “Fine. But if that kid on the Big Wheel even looks at me sideways, I’m deploying the cricket contingency.” “Noted. Now let’s go. We need to buy a present to put in the big empty gift bag so we look like we belong.” Bill’s face lit up. “A Trojan horse! Now *that’s* good tradecraft.”
created_at
"2026-02-16T21:32:30.139377+00:00"
2/16/2026, 9:32:30 PM
model
"deepseek/deepseek-chat-v3.1"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
1430
moderation_latency
(null)
generation_time
105050
tokens_prompt
95
tokens_completion
1005
native_tokens_prompt
67
native_tokens_completion
914
native_tokens_completion_images
(null)
native_tokens_reasoning
0
native_tokens_cached
17
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
0.0007347700000034
router
(null)
provider_responses
0
id
"cmpl-c09be4914dab4ea8966cbe0f7bcfa60a"
status
200
is_byok
false
latency
68
endpoint_id
"012f7d1f-23d4-4d03-9aa3-432d8ad82757"
provider_name
"DeepInfra"
model_permaslug
"deepseek/deepseek-chat-v3.1"
api_type
"completions"
id
"gen-1771277444-6GPzwTR9m7aPtG6T7OIi"
upstream_id
"cmpl-c09be4914dab4ea8966cbe0f7bcfa60a"
total_cost
0.0007347700000034
cache_discount
0.0000013599999966
upstream_inference_cost
0
provider_name
"DeepInfra"
Evaluation details
Result
Evaluator
Details
Meta Data
0.0000%
Matches word count
n/a
neededClean
false
words
611
7.2192%
Dialogue to Total Word Ratio
Ratio: 77.25%, Deviation: 7.25%
neededClean
false
wordsTotal
611
wordsDialogue
472
3.6096%