NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-10-01M25DN7A0BMDES3DYKRT5FN7Z
Run Details
Model:
DeepSeek V4.1 Flash (Reasoning, High)
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
42.0s
Cost:
$0.006439
Provider:
DeepSeek
Evaluation summary
Score
Evaluator
Details Link
90.48%
Matches word count
(details)
99.77%
Dialogue to Total Word Ratio
(details)
95.1271%
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
Molly unfolded the crayon-smeared invitation on the hood of their van and weighed it down with a lug wrench. The paper announced Timothy’s Superhero Space Dinosaur Party, Saturday at two, Community Center Room B, with a glittery postscript: Presents welcome. To most people, that was a cheerful invitation. To Molly, it was a vault with cake. She had spent three weeks casing the community center. Its security consisted of a teenager named Kyle, a locked cabinet, and a sign asking visitors not to feed the raccoons. Kyle took smoke breaks precisely every forty minutes. The raccoons were the wildcard. Bill hated raccoons. He hated Kyle, though they had never met. It was a matter of principle and dental floss. Bill sat on a milk crate, cleaning a stethoscope he had bought for a medical heist that had collapsed when the patient woke up. He squinted at the invitation. “Children’s party,” he said. “That’s a new low.” “It’s not low. It’s tactical.” Molly tapped the gift table diagram she had drawn from memory. “Affluent grandparents. Theme parties. Unattended presents.” “We’re robbing a seven-year-old.” “We’re liberating corporate merchandise from a juvenile redistribution hub.” “That’s worse.” Molly ignored him. The plan required two performers, one distraction, and a diaper bag large enough to hold a drone, three gift cards, and whatever foolishness came in a velvet box. Bill would be the clown. Molly would be Princess Nebula, a character she had invented that morning and already disliked. “Why am I the clown?” Bill asked. “Because you have sad eyes. Children trust sad eyes.” “I have a criminal record.” “So do clowns.” Bill stood and paced beside the van. Across the street, a birthday supply store displayed piñatas shaped like unicorns and minor celebrities. He hated that he was already calculating their resale value. “What about the kids?” he said. “Kids don’t care about presents once the cake appears. Cake is a powerful narcotic.” “And the parents?” “Exhausted. Filming each other. We avoid the Warden.” “The Warden?” “Cargo shorts. Zip ties. Volunteers too hard.” Bill nodded grimly. Every party had one: a dad with a clipboard and a hero complex. “Exit?” he asked. “Kitchen. I start a small fire in the trash can.” “Controlled?” “Mostly.” “Molly.” “Controlled enough.” He rubbed his temples. “How much are we talking?” “Enough to pay off the van and buy new identities.” “I like Bill.” “You like being Bill?” “It’s easy to spell.” Molly folded the invitation and slid it into her jacket. She handed Bill a bag of balloons and a practice pump. He stared at them as if they were evidence. “I need a balloon dog,” she said. “It looks like a snake with tumors.” “Then call it a space snake. Improvise.” They climbed into the van. Bill practiced knotting balloons while Molly ran the stop signs of a route she had memorized. At two o’clock, they would enter as entertainers, smile at strangers, and leave with a diaper bag full of birthday miracles. It was small, glittery, and profoundly illegal. In other words, a perfect afternoon.
created_at
"2026-09-10T10:26:18.054Z"
9/10/2026, 10:26:18 AM
model
"deepseek/deepseek-v4.1-flash-20260910"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
451
moderation_latency
(null)
generation_time
41912
tokens_prompt
75
tokens_completion
10240
native_tokens_prompt
93
native_tokens_completion
10709
native_tokens_completion_images
(null)
native_tokens_reasoning
9987
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"
service_tier
(null)
usage
0.00643935
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.20.0; linux; x64))"
http_referer
(null)
request_id
"req-1789035978-vXO8NkK40yuzpSttbc1g"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1789035978-lsBjfYgFpeDN64MVDt1a"
upstream_id
"f07149fd-4415-4031-b61a-60ddbcc35e0a"
provider_responses
0
endpoint_id
"fd104bc8-535a-4d31-a847-a8e402e86a02"
id
"f07149fd-4415-4031-b61a-60ddbcc35e0a"
is_byok
false
latency
335
model_permaslug
"deepseek/deepseek-v4.1-flash-20260910"
provider_name
"DeepSeek"
status
200
total_cost
0.00643935
cache_discount
(null)
upstream_inference_cost
0
provider_name
"DeepSeek"
response_cache_source_id
(null)
data_region
"global"
workspace_id
"97e315e5-d303-487d-83c1-83180e8a13d4"
Evaluation details
Result
Evaluator
Details
Meta Data
90.48%
Matches word count
n/a
neededClean
false
words
510
99.77%
Dialogue to Total Word Ratio
Ratio: 30.21%, Deviation: 0.21%
neededClean
false
wordsTotal
513
wordsDialogue
155
95.1271%