NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-04-25-01KQ2H1J9HN0296TS8445BRB87
Run Details
Model:
DeepSeek V4 Flash (Reasoning)
Test:
Dialogue tags
Scenario:
Write 500 words with 50% dialogue
Duration:
13.0s
Cost:
$0.000251
Provider:
DeepSeek
Evaluation summary
Score
Evaluator
Details Link
0.00%
Matches word count
(details)
0.00%
Dialogue to Total Word Ratio
(details)
0.0000%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 500 word scene that has 50% 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
“Bill, I’m telling you, this is our golden ticket. Kids’ parties are the new banks. Parents drop hundreds on toys, and nobody guards a bouncy castle.” Molly adjusted her ski mask, which was too small and kept riding up her nose. Bill stared at the invitation scrawled on unicorn stationery. “We’re breaking into a seven-year-old’s birthday. At a petting zoo. With a magician named Professor Sparkles. I’m supposed to be afraid of a man who pulls rabbits out of hats?” “You should be. Rabbits have teeth. And he might have a wand. That’s basically a weapon if you’re not careful.” “Molly, we’ve robbed an armored car. We once lifted a Fabergé egg while the owner was in the shower. This is low-hanging fruit. Literally. The tote bags are labeled ‘Rainbow Rainbow’ and they’re hanging on pegs.” “Exactly. Low expectations, low security. The parents are too busy filming their little darlings eating cake to watch the gift table. We walk in, we look like aunt and uncle, we grab the haul. In and out before the piñata gets whacked.” Bill rubbed his temples. “We need disguises. I’m not wearing a unicorn onesie again. Last time the horn got caught in the fire escape.” “No onesies. I’ve got a plan. We pose as balloon artists. You twist a poodle, I distract with a giraffe. The kids love that, parents get weepy, and while they’re clapping, we swap the gift bags with identical empty ones we brought.” “Empty bags? So we’re stealing presents by giving them nothing? That’s just a magic trick.” “More like a confidence trick. There’s fifty kids. Each bag has at least fifty bucks worth of junk. That’s two and a half grand, tax-free. And no one suspects the balloon lady.” Bill sighed. “And what do we do with a hundred plastic dinosaurs and half-eaten candy bars?” “We fence them. There’s a guy in a strip mall who buys unopened toys. No questions. He once took a talking Elmo that swore in Spanish. Didn’t blink.” “I’m not sure I want to be the guy who steals from kindergarteners. It feels… low.” Molly poked him in the chest. “Low is what pays the bills. Remember last week when we tried to lift a Ming vase from a museum and ended up stealing a coat rack? This is a sure thing.” “Fine. But I’m not doing the balloon-animal voice. That squeaky thing you do creeps me out.” “Deal. I’ll do the voice. You do the hands. Now, the party starts at 2. We arrive at 1:45, set up our station near the gift table. When the magician starts his show, everyone looks at him. That’s our window. We switch bags, stuff ours into a duffel, and walk out during the finale.” “And if someone asks for a balloon swan?” “Then you make a swan. Stop being difficult. This is art. Theft is art. And children’s parties are the new frontier. Untapped. Overlooked. Profitable.” Bill looked at the invitation again. “It’s at ‘Pony Paradise.’ There’s a sign that says ‘No Crying Allowed.’ We are going to make a lot of people cry.” Molly grinned. “That’s the spirit. Now, let’s practice. I’ll be Professor Sparkles. You be the terrified dad. Action.”
created_at
"2026-04-25T14:35:04.885Z"
4/25/2026, 2:35:04 PM
model
"deepseek/deepseek-v4-flash-20260423"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
526
moderation_latency
(null)
generation_time
12634
tokens_prompt
77
tokens_completion
926
native_tokens_prompt
67
native_tokens_completion
864
native_tokens_completion_images
(null)
native_tokens_reasoning
78
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
0.0002513
router
(null)
provider_responses
0
endpoint_id
"934a69f9-bd54-474b-beca-24560f721e12"
is_byok
false
latency
40
model_permaslug
"deepseek/deepseek-v4-flash-20260423"
provider_name
"DeepInfra"
status
429
1
endpoint_id
"722e0746-2034-4d6e-a08f-e6271849ee9e"
id
"a9b2b1d9-b1ab-4d29-a4fa-7e52572c6d9d"
is_byok
false
latency
275
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-1777127704-wN2Ced2YeIooC9mJC7xx"
session_id
(null)
api_type
"completions"
id
"gen-1777127704-YD3bebHFUgGCMRlpt12M"
upstream_id
"a9b2b1d9-b1ab-4d29-a4fa-7e52572c6d9d"
total_cost
0.0002513
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
537
0.00%
Dialogue to Total Word Ratio
Ratio: 91.90%, Deviation: 41.90%
neededClean
false
wordsTotal
543
wordsDialogue
499
0.0000%