NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-18-01KM07GKCDKCMDK4CA6JRCPD6R
Run Details
Model:
Mistral Small 4
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
5.5s
Cost:
$0.000357
Provider:
Mistral
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
0.00%
Dialogue to Total Word Ratio
(details)
0.0000%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 30% 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 rain drummed against the window of the stolen van, its rhythmic patter the only sound in the quiet street as Molly and Bill hunched over the stolen floor plan. The house they were targeting was a two-story monstrosity on the cul-de-sac, its windows glowing with the sickly-sweet glow of birthday decorations. A tacky "Happy 7th Birthday, Tommy!" banner hung crookedly across the garage door, the letters flapping slightly in the breeze. Molly traced a grease-stained finger along the blueprint, her brow furrowed. "Alright, Bill, listen up. The parents are throwing some big shindig for this kid—balloons, clown, the works. But get this—they stashed the presents in the guest room upstairs. Kid’s got, like, three iPads already." Bill scratched his stubble, his salt-and-pepper beard matted with last night’s leftover donut crumbs. "Three iPads? Who gives a seven-year-old three iPads?" Molly shrugged. "Rich people problems. Anyway, security’s a joke. Lame security system you can bypass with a banana and a paperclip, but they hired some half-asleep teen to watch the door. College kid named Chad—plays video games the whole time." Bill grinned, popping a breath mint into his mouth. "So we’re good, then. We go in, grab the loot, and bounce before anyone notices the bacon we’ll stage as a fire hazard." Molly shot him a look. "The bacon? Really?" "Works every time," Bill insisted. "Nothing panics people like the smell of burnt meat." She sighed, rolling up the blueprint. "Fine. But we need to time this right. The clown’s gonna perform at seven, and all the parents will be in the living room laughing at a guy in a rainbow wig. Complete and utter chaos. That’s our window." Bill nodded, grabbing a handful of tools from his duffel bag. "Perfect. We go in at seven-fifteen. Plenty of time to crack the window, disable the camera, and snag the presents before the parents even realize Tommy’s gifts are gone." Molly checked her watch. "We’ve got two hours. You got the distraction ready?" "Oh, I’m ready," Bill said, pulling a kazoo out of his pocket and giving it a triumphant blow. She groaned. "You’re not seriously using sound again, are you?" "What?” Bill feigned innocence. “It’s foolproof. People lose their minds over random noises in a birthday party. Last job, I mimicked a fire alarm and got away with a year’s supply of juice boxes."
created_at
"2026-03-18T10:22:57.431Z"
3/18/2026, 10:22:57 AM
model
"mistralai/mistral-small-2603"
1/1/2603, 12:00:00 AM
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
384
moderation_latency
(null)
generation_time
5390
tokens_prompt
77
tokens_completion
624
native_tokens_prompt
81
native_tokens_completion
575
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
0.00035715
router
(null)
provider_responses
0
endpoint_id
"f5ed6283-4e17-457d-98cd-2ec0c318171d"
id
"b91eb7b898304dc799393688b3ed8589"
is_byok
false
latency
383
model_permaslug
"mistralai/mistral-small-2603"
1/1/2603, 12:00:00 AM
provider_name
"Mistral"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.14.0; linux; x64))"
http_referer
(null)
api_type
"completions"
id
"gen-1773829377-ancku7jov50BPCQP9LLN"
upstream_id
"b91eb7b898304dc799393688b3ed8589"
total_cost
0.00035715
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Mistral"
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Matches word count
n/a
neededClean
false
words
396
0.00%
Dialogue to Total Word Ratio
Ratio: 60.93%, Deviation: 30.93%
neededClean
false
wordsTotal
407
wordsDialogue
248
0.0000%