NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-04-25-01KQ2EZ733V36F3GP25H1BGQQ0
Run Details
Model:
DeepSeek V4 Pro (Reasoning)
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
6m 15s
Cost:
$0.015144
Provider:
DeepSeek
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
99.94%
Dialogue to Total Word Ratio
(details)
49.9698%
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 spread the blueprints across the sticky diner table, beside a syrup-smeared placemat decorated with smiling pancakes. The blueprints were actually a hand-drawn map of a suburban backyard, complete with a bouncy castle labeled “Fort Knox” and a gift table marked “LOOT.” Bill squinted at the cartoon unicorns and the crayon arrows indicating wind direction. A half-eaten stack of pancakes sat between them, growing cold. Bill tapped a glitter-pen marking. “You want us to dress as clowns.” Molly beamed, producing two rainbow wigs from a duffel bag that smelled faintly of old balloons. “Not just clowns, Bill. Sparkles and Giggles, from the Joyful Jingles Agency. I’ve already booked us.” Bill picked up a wig with two fingers, as if it might trigger an allergy. The synthetic fibers crackled with static. He remembered a childhood incident involving a unicycle and a llama, and shuddered. “The shoes are a dealbreaker.” Molly ignored him. She unfolded a secondary diagram illustrating the trajectory of glitter-filled hamsters. The hamsters were drawn with tiny capes. She had spent six weeks training them to scatter on command, using a Pavlovian method involving carrot sticks and a kazoo. “Diversion. You’ll do a magic trick where I vanish into the gift pile.” Bill stared at the drawing of a stick figure disappearing behind a three-tiered cake. The cake had googly eyes. He wondered if any actual children’s parties had this level of planning. “I can’t even find the hope in a Kinder Egg.” Molly slid a pair of floppy shoes across the table. They were size fourteen, and one squeaked involuntarily. “You’ll just honk and wave a wand. The hamsters handle the chaos.” Bill raised an eyebrow. “Where do you get glitter-covered hamsters on short notice?” “A man called Nibbles. He owes me a favor.” Bill sighed, fogging the sugar pourer. He thought about the moral implications, but the thought was interrupted by a waitress refilling their coffee. The waitress had a name tag that read “Nibbles,” which felt like an omen. Bill leaned in. “What’s the getaway?” “Ice-cream truck. Plays the Baby Shark remix. No pursuit.” “And the moral quandary of stealing from tiny humans?” “We’re not stealing joy, just relocating it. We’ll mail each child a handcrafted apology note. On glitter paper.” Bill nodded slowly, a man accepting a ridiculous fate. “Fine. But if I get a honk-related injury, I’m billing you.” Outside the diner window, a pigeon fought a discarded balloon animal. The scene felt like a metaphor. Bill imagined the presents: LEGO sets, dolls that peed, video games. He could already hear the sirens, or maybe just the endless loop of Baby Shark. He took a deep breath and grabbed the wig.
created_at
"2026-04-25T13:58:50.728Z"
4/25/2026, 1:58:50 PM
model
"deepseek/deepseek-v4-pro-20260423"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
881
moderation_latency
(null)
generation_time
374642
tokens_prompt
77
tokens_completion
15815
native_tokens_prompt
67
native_tokens_completion
17373
native_tokens_completion_images
(null)
native_tokens_reasoning
16737
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
0.015143655
router
(null)
provider_responses
0
endpoint_id
"d084170b-dd99-4e53-9ae1-b63881f3cfbf"
id
"7dba962e-a7fe-4c59-8c7f-4baf2e55db57"
is_byok
false
latency
289
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-1777125530-NKCxxQr7ZSWbKVIH96Uq"
session_id
(null)
api_type
"completions"
id
"gen-1777125530-dQDaG67yakXTaxjNHjvb"
upstream_id
"7dba962e-a7fe-4c59-8c7f-4baf2e55db57"
total_cost
0.015143655
cache_discount
(null)
upstream_inference_cost
0
provider_name
"DeepSeek"
response_cache_source_id
(null)
Evaluation details
Result
Evaluator
Details
Meta Data
0.00%
Matches word count
n/a
neededClean
false
words
447
99.94%
Dialogue to Total Word Ratio
Ratio: 30.11%, Deviation: 0.11%
neededClean
false
wordsTotal
455
wordsDialogue
137
49.9698%