NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-04-25-01KQ2H3RAPTJF7YVZ69TB6XRWG
Run Details
Model:
DeepSeek V4 Pro
Test:
Dialogue tags
Scenario:
Write 200 words with 10% dialogue
Duration:
7.7s
Cost:
$0.001505
Provider:
Together
Evaluation summary
Score
Evaluator
Details Link
3.62%
Matches word count
(details)
0.00%
Dialogue to Total Word Ratio
(details)
1.8117%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 200 word scene that has 10% 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 planning stage, Molly felt, should not involve sitting in a van that smelled of old carpet and Bill’s egg sandwich. A street away, the bouncy castle listed in a suburban garden like a collapsed lung. Through the windscreen, they watched a man in a waistcoat three sizes too small attempt balloon animals. He was currently producing what looked like a distressed sausage. Molly lowered her binoculars. “The clown is a variable,” she said. Bill, consulting a blueprint he’d crayoned onto a napkin, didn’t look up. “He’s not a variable, he’s a man called Terry who smells of damp. The real variable is the pass the parcel music stopping, triggering a gift-grab reflex in the under-fives. That’s our sonic event horizon.” “Your understanding of children is terrifyingly academic,” Molly said. “Remember, you’re the bumbling wizard’s assistant. Your sleeve needs to swallow multiple presents. The birthday girl’s mother already thinks my catering assistant cover is genuine.” Bill finally glanced at the garden, where the clown tripped over a Labrador. “We’re pilfering a mountain of brightly coloured plastic, Mary. The loot bags alone will have enough sugar to fuel a small car. Our exit strategy depends entirely on you dropping a tray of jelly at the right moment.” “A classic,” Molly agreed. “Distraction by structural failure. Let’s just pray the magician doesn’t pull a better trick.”
created_at
"2026-04-25T14:36:16.605Z"
4/25/2026, 2:36:16 PM
model
"deepseek/deepseek-v4-pro-20260423"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
718
moderation_latency
(null)
generation_time
7334
tokens_prompt
77
tokens_completion
376
native_tokens_prompt
67
native_tokens_completion
310
native_tokens_completion_images
(null)
native_tokens_reasoning
0
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
0.0015047
router
(null)
provider_responses
0
endpoint_id
"d084170b-dd99-4e53-9ae1-b63881f3cfbf"
is_byok
false
latency
271
model_permaslug
"deepseek/deepseek-v4-pro-20260423"
provider_name
"DeepSeek"
status
429
1
endpoint_id
"13bda129-f8a6-4a60-b1db-98082e5f7ccd"
id
"ogZhNt3-2dTqGa-9f1e143d838ca249"
is_byok
false
latency
718
model_permaslug
"deepseek/deepseek-v4-pro-20260423"
provider_name
"Together"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.14.1; linux; x64))"
http_referer
(null)
request_id
"req-1777127776-g13iy9skAetQmqTE4gsc"
session_id
(null)
api_type
"completions"
id
"gen-1777127776-WcivLB7w4Pmlv0SVJL3T"
upstream_id
"ogZhNt3-2dTqGa-9f1e143d838ca249"
total_cost
0.0015047
cache_discount
(null)
upstream_inference_cost
0
provider_name
"Together"
response_cache_source_id
(null)
Evaluation details
Result
Evaluator
Details
Meta Data
3.62%
Matches word count
n/a
neededClean
false
words
224
0.00%
Dialogue to Total Word Ratio
Ratio: 56.64%, Deviation: 46.64%
neededClean
false
wordsTotal
226
wordsDialogue
128
1.8117%