NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-09-03-01M1KBE4DXH87N3S94V74KGTXM
Run Details
Model:
DeepSeek V4 Pro 0813 (Reasoning, High)
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
2m 4s
Cost:
$0.035679
Provider:
StreamLake
Evaluation summary
Score
Evaluator
Details Link
100.00%
Matches word count
(details)
92.63%
Dialogue to Total Word Ratio
(details)
96.3131%
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 crinkled map of the community center across the sticky diner table, narrowly missing Bill's coffee. She tapped a purple fingerprint smudge near the bouncy castle room and waited for him to stop chewing his pancake like a man reconsidering every choice that had led him to breakfast with a professional thief. The map had been drawn in crayon by a five-year-old informant, and the unicorn in the corner was, according to Molly, a scale model of the gift table. Bill didn't ask why she trusted a kindergartner with surveillance. He had learned that questions only lengthened the planning phase. Bill finally swallowed. "You're sure the loot is in the Rainbow Room?" Molly nodded. "Mountain of glittery boxes, Bill. Three LEGO sets minimum." He poked the map with his fork. "No pony. I draw the line at livestock." "No pony. Just toys, gift cards, and a fog machine magician named Wacky Dave." She explained the rest while drawing an X over the side door. Bill would rent the inflatable dinosaur costume, the one with the broken tail, and stagger around outside the window during the second hour, when the sugar crash hit and parents chased toddlers. Molly would slip inside dressed as a party clean-up volunteer, clip-on badge already laminated. Bill frowned at a syrup stain shaped like a unicorn. He tried to remember which of his three ex-girlfriends had called him emotionally a bouncy castle. "Why am I always the dinosaur?" "Long, panicky, getaway legs." He sighed and wrote something in a tiny notebook. "Fine. But I'm not roaring. A kid threw a juice box at my snout last time." Molly leaned forward, lowering her voice. "The assistant is a former mall cop. She smells like cotton candy and suspicion." "Suspicion has a smell?" "Burnt sugar and cheap perfume." Bill took a slow sip of coffee and considered the bouncy castle emergency flap, the parking lot, the van idling behind a recycling bin. Molly's plan had the usual mix of absurdity and glue-gun craft supplies. The badge said Official Gift Checker. It was laminated. He could not argue with lamination. "And if a child sees you?" "I'm the present inspector. Confidence and a clip-on tie." "You're wearing a clip-on tie to a children's party?" "Professionalism, Bill. It matters." He set his mug down. "If there's a piñata, I'm out. I still have nightmares from the donkey job in '09." "No piñata. Wacky Dave does a live rabbit. Very distracting." Bill's face softened. "I like rabbits." "Bond with it while I commit grand larceny. Finish your pancakes." Molly slid the map toward her, folded it into a paper airplane, and launched it at the napkin dispenser. It nose-dived into syrup. They sat in silence for a moment, watching the unicorn dissolve into a sticky rainbow. Outside the diner window, a toddler in a tiara staggered past, screaming with joy, and Bill took it as a sign. He then left a small five-dollar tip because even criminals had standards.
created_at
"2026-09-03T10:01:08.812Z"
9/3/2026, 10:01:08 AM
model
"deepseek/deepseek-v4-pro-20260813"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
1155
moderation_latency
(null)
generation_time
124055
tokens_prompt
75
tokens_completion
10191
native_tokens_prompt
146
native_tokens_completion
10614
native_tokens_completion_images
(null)
native_tokens_reasoning
9939
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
0.035679415
router
(null)
user_agent
"langchainjs-openai/1.0.0 ((node/v24.19.0; linux; x64))"
http_referer
(null)
request_id
"req-1788429668-6WJpd7nUppJdnvT6wJRh"
session_id
(null)
preset_id
(null)
api_type
"completions"
id
"gen-1788429668-csrzvli9mXnIkWAZqUPx"
upstream_id
"req-wx80rn-1788429668976725800"
provider_responses
0
endpoint_id
"072c1c23-c069-4242-bb6d-1b30ca5f4eb8"
id
"req-wx80rn-1788429668976725800"
is_byok
false
latency
1155
model_permaslug
"deepseek/deepseek-v4-pro-20260813"
provider_name
"StreamLake"
status
200
total_cost
0.035679415
cache_discount
(null)
upstream_inference_cost
0
provider_name
"StreamLake"
response_cache_source_id
(null)
data_region
"global"
workspace_id
"97e315e5-d303-487d-83c1-83180e8a13d4"
Evaluation details
Result
Evaluator
Details
Meta Data
100.00%
Matches word count
n/a
neededClean
false
words
500
92.63%
Dialogue to Total Word Ratio
Ratio: 31.24%, Deviation: 1.24%
neededClean
false
wordsTotal
509
wordsDialogue
159
96.3131%