NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-04-25-01KQ2GSSA3C6GHF2X9ZHP1XK6H
Run Details
Model:
DeepSeek V4 Flash (Reasoning)
Test:
Dialogue tags
Scenario:
Write 500 words with 30% dialogue
Duration:
13.6s
Cost:
$0.000238
Provider:
DeepSeek
Evaluation summary
Score
Evaluator
Details Link
95.99%
Matches word count
(details)
0.00%
Dialogue to Total Word Ratio
(details)
47.9934%
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 van idled at the curb, its engine a low grumble against the sugary shrieks drifting from the backyard. Molly adjusted the clown wig, the orange curls bouncing absurdly around her face. “You look like a melting satsuma,” Bill said, not looking up from the tablet. “And you look like a man who’s about to be jumped by a five-year-old with a juice box. Did you get the balloon animals?” “I got a pump and twelve feet of latex. We are not making animals. We are making a distraction.” Molly peered through the fence slats. A bounce castle swelled in the afternoon sun, a pink and yellow tumor of joy. A piñata in the shape of a unicorn swung from a tree branch, already beaten to a pulp. “The loot table is by the gazebo. Wrapped in cartoon giraffe paper. Tacky.” “Focus. The parents are inside, doing that ‘adult conversation’ thing. The kids are feral. We have a three-minute window when the magician starts his act.” Bill swiped on his tablet, bringing up a schematic. “I enter from the side gate, you create chaos with the balloons. A ruptured bladder, a loud pop, everyone looks. I grab the bags, we meet at the back of the shed.” “And if the magician asks for a volunteer?” “Then you go up and get sawed in half. That’s the plan.” Molly grinned, her teeth too white against the clown makeup. “You’re all heart, Bill. But what about the present pile? That’s twenty gifts, minimum. You expect to carry that without looking like Santa’s drunk cousin?” “I have a laundry bag. Canvas. Unobtrusive.” “Unobtrusive. Right. Because a man in a black turtleneck stuffing presents into a sack during a children’s party is just blending in with the scenery.” “I’m wearing a polo shirt. It’s light blue.” “Oh, my mistake. Light blue. That changes everything. They’ll think you’re the fun uncle.” The shrieks from the backyard peaked, then dropped. The magician’s muffled voice boomed over a portable speaker. “And now, for my next trick—a lovely white dove!” “That’s the cue,” Bill said, pocketing the tablet. He grabbed the balloon pump and a fistful of latex. “Remember—three minutes. Then we’re gone.” Molly cracked her neck, the wig slipping. “Three minutes. Piece of cake. Or should I say, piece of cake smash? Because that’s what we’re about to do to their party.” Bill paused at the gate. “You’re enjoying this too much.” “Bill,” she said, her voice dropping flat, “they asked for a unicorn piñata. A unicorn. That’s for a girl named Olivia. I saw the banner. The theme is ‘Princess and Dragons.’ They have tiny plastic crowns. I’m not enjoying this. I’m cleansing the world of delusional optimism.” “Just pop the balloon, Molly.” “I’ll pop more than that.” She winked, but the clown paint made it look like a threat. Bill sighed, pushed open the gate, and stepped into the party. The magician’s dove flapped up into a tree. Chaos was about to be served, with a side of plastic jewels and candy.
created_at
"2026-04-25T14:30:49.927Z"
4/25/2026, 2:30:49 PM
model
"deepseek/deepseek-v4-flash-20260423"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
690
moderation_latency
(null)
generation_time
13547
tokens_prompt
77
tokens_completion
899
native_tokens_prompt
67
native_tokens_completion
818
native_tokens_completion_images
(null)
native_tokens_reasoning
86
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
0.00023842
router
(null)
provider_responses
0
endpoint_id
"722e0746-2034-4d6e-a08f-e6271849ee9e"
id
"3b9a04ab-3f10-43cf-84ac-694f4984c2f0"
is_byok
false
latency
354
model_permaslug
"deepseek/deepseek-v4-flash-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-1777127449-KnVpW4n3QeCFa1QuGgKW"
session_id
(null)
api_type
"completions"
id
"gen-1777127449-P5jiTdSJqbu3AMSKRPBV"
upstream_id
"3b9a04ab-3f10-43cf-84ac-694f4984c2f0"
total_cost
0.00023842
cache_discount
(null)
upstream_inference_cost
0
provider_name
"DeepSeek"
response_cache_source_id
(null)
Evaluation details
Result
Evaluator
Details
Meta Data
95.99%
Matches word count
n/a
neededClean
false
words
508
0.00%
Dialogue to Total Word Ratio
Ratio: 62.23%, Deviation: 32.23%
neededClean
false
wordsTotal
511
wordsDialogue
318
47.9934%