NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-04-25-01KQ2JPNEV6R1D2X4EWP4MEBZK
Run Details
Model:
DeepSeek V4 Pro
Test:
Dialogue tags
Scenario:
Write 200 words with 10% dialogue
Duration:
34.6s
Cost:
$0.000293
Provider:
DeepSeek
Evaluation summary
Score
Evaluator
Details Link
68.10%
Matches word count
(details)
56.07%
Dialogue to Total Word Ratio
(details)
62.0861%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 200 word scene that has 10% 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 unspooled a fire-escape blueprint across the sticky surface of the booth’s table, right next to a desiccated puddle of maple syrup. Bill leaned in, his gaze darting across the annotations as if they were scripture smuggled out of a particularly saccharine Vatican. Molly tapped a circle marked ‘Bouncy Castle Egress’ with a plastic cocktail sword. “Security’s a clown, a magician, and a teenager who’s already cried twice,” she said. Bill nodded gravely, extracting a stray bit of glitter from his ear. “The magician’s the real threat,” he replied. “Never trust a man who can palm your wallet and a dove simultaneously.” The plan was elegant: Molly would pose as a disgraced fairy-tale princess peddling gluten-free face paint while Bill, crammed into a rented unicorn costume, manufactured a hydraulic failure on the cake table to create a frosting-based diversion. “And the piñata?” Bill asked, his voice muffled as he practiced breathing through the costume’s single, suffocating nostril. Molly flipped the page, revealing a schematic of a blindfolded child swinging at a candy-filled decoy while they employed a decoy child to misfire a party popper into the clown’s wig. The real target, a monolithic gleaming fortress of tiered presents, sat in the venue’s bay window visible only from a crawlspace accessible through a ball pit.
created_at
"2026-04-25T15:04:31.104Z"
4/25/2026, 3:04:31 PM
model
"deepseek/deepseek-v4-pro-20260423"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
1450
moderation_latency
(null)
generation_time
8298
tokens_prompt
77
tokens_completion
349
native_tokens_prompt
67
native_tokens_completion
303
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)
num_fetches
(null)
web_search_engine
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"stop"
usage
0.000292755
router
(null)
provider_responses
0
endpoint_id
"d084170b-dd99-4e53-9ae1-b63881f3cfbf"
id
"87a6d14a-373a-49ce-8473-db00b6a35ce5"
is_byok
false
latency
420
model_permaslug
"deepseek/deepseek-v4-pro-20260423"
provider_name
"DeepSeek"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.14.1; linux; x64))"
http_referer
(null)
request_id
"req-1777129471-RrG4a2RCoqGmQGz5p3IL"
session_id
(null)
api_type
"completions"
id
"gen-1777129471-QfqBSJt1VwjCEQjWVRfu"
upstream_id
"87a6d14a-373a-49ce-8473-db00b6a35ce5"
total_cost
0.000292755
cache_discount
(null)
upstream_inference_cost
0
provider_name
"DeepSeek"
response_cache_source_id
(null)
Evaluation details
Result
Evaluator
Details
Meta Data
68.10%
Matches word count
n/a
neededClean
false
words
214
56.07%
Dialogue to Total Word Ratio
Ratio: 18.72%, Deviation: 8.72%
neededClean
false
wordsTotal
219
wordsDialogue
41
62.0861%