NC Bench
Overview
About
Tests
Models
Vendors
Run: 2026-02-16-01KHK9VSC6JHERV0W7GWXPPF2Y
Run Details
Model:
GPT-5.2
Test:
Dialogue tags
Scenario:
Write 200 words with 90% dialogue
Duration:
43.1s
Cost:
$0.036936
Provider:
OpenAI
Model input
User/Human
You are a author of comedic heist books. Please write a 200 word scene that has 90% 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
They crouched behind the bounce house, watching candlelight flicker inside. "Bill, remember: we’re not stealing joy; we’re repossessing surplus." "Molly, it’s a six-year-old’s birthday. The surplus is hugs." "Hugs don’t fit in my duffel. Presents do." "How do we get past the clown? He’s built like a refrigerator." "Easy. We wear party hats. Nobody questions hats." "I’m six-foot-four, Molly." "Then you’re a very tall six-year-old with seasonal allergies." "And the loot plan?" "You distract the parents by asking about gluten-free cake." "They’ll smell crime on me." "No, Bill, they’ll smell desperation for dietary validation." "While they argue, I slide under the gift table." "Under? You hate crawling." "I hate prison more." "What about the birthday kid?" "We gift him a decoy: an empty box labeled ‘Imagination.’" "That’s evil." "It’s educational." "And exit route?" "Through the piñata line. Everybody’s blindfolded." "If someone grabs me?" "Say you’re the donkey. Bray convincingly." "I can’t bray." "Bill, for this score, you will learn." "Also, pocket thank-you cards; they’re pre-addressed for fencing." "Who buys thank-you cards?" "Grandmothers, Bill. They’ll pay anything for politeness." "And if the cops show?" "We vanish into the magician’s smoke machine." "I rented it. Crime is just theater with worse reviews."
created_at
"2026-02-16T13:23:14.061377+00:00"
2/16/2026, 1:23:14 PM
model
"openai/gpt-5.2-20251211"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
8852
moderation_latency
(null)
generation_time
43030
tokens_prompt
77
tokens_completion
992
native_tokens_prompt
66
native_tokens_completion
2630
native_tokens_completion_images
0
native_tokens_reasoning
2304
native_tokens_cached
0
num_media_prompt
(null)
num_input_audio_prompt
(null)
num_media_completion
0
num_search_results
(null)
origin
"https://ncbench.com/"
is_byok
false
finish_reason
"stop"
native_finish_reason
"completed"
usage
0.0369355
router
(null)
provider_responses
0
id
"resp_0e475aa1896915330169931a15fc348194a6876d48ebee9b66"
status
200
is_byok
false
latency
173
endpoint_id
"f00142c2-6a93-49ce-9e36-5593b904ce3b"
provider_name
"OpenAI"
model_permaslug
"openai/gpt-5.2-20251211"
api_type
"completions"
id
"gen-1771248149-V0inheBDyvdCo9bN3BMC"
upstream_id
"resp_0e475aa1896915330169931a15fc348194a6876d48ebee9b66"
total_cost
0.0369355
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenAI"
Evaluation details
Result
Evaluator
Details
Meta Data
93.6496%
Matches word count
n/a
neededClean
false
words
209
92.8689%
Dialogue to Total Word Ratio
Ratio: 95.22%, Deviation: 5.22%
neededClean
false
wordsTotal
209
wordsDialogue
199
93.2593%