NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-09-01KKAC687QAH5MK80DGXMDEA43
Run Details
Model:
GPT-5.4 (Reasoning, Low)
Test:
Dialogue tags
Scenario:
Write 200 words with 90% dialogue
Duration:
10.5s
Cost:
$0.005580
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
68.10%
Matches word count
(details)
98.43%
Dialogue to Total Word Ratio
(details)
83.2647%
Overall
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
Molly peered through the hedge while Bill unfolded a crayon-smeared map. “Right, Bill, target acquired: one bouncy castle, one distracted magician, and approximately forty gifts wrapped by amateurs.” “Amateurs with ribbon, Molly. Ribbon is the burglar’s welcome mat.” “Focus. Entry point?” “Cake table flank. No one guards the cake until the candles appear.” “Counterpoint: sticky icing. I refuse to die with buttercream on my cuffs.” “Then we go in as entertainers.” “You can’t juggle.” “I can drop things with confidence.” “Children can smell fear.” “So can poodles, and yet I once robbed a vet.” “We need disguises. Balloon artist and depressed pirate?” “I’m not doing balloons. They squeak like guilty mice.” “Fine. Face painter and uncle nobody recognizes.” “I was born to be an unconvincing uncle.” “Extraction route?” “Through the garden gate, past the inflatable palm tree, into the alley, presents into the van.” “The van still says ‘Drain Wizard.’” “Excellent camouflage. Nobody suspects a plumber with a sack.” “What about the parents?” “They’ll be filming a child attacking a piñata.” “What about the dog?” “I’ve brought sausages.” “What about the clown?” “I’ve brought two sausages.” “Molly, this is reckless.” “Bill, it’s a six-year-old’s birthday party. We are competing with juice boxes.” Bill sighed. “Then let’s rob childhood before the magician finds his dove.”
created_at
"2026-03-09T22:41:23.711Z"
3/9/2026, 10:41:23 PM
model
"openai/gpt-5.4-20260305"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
3818
moderation_latency
(null)
generation_time
10399
tokens_prompt
77
tokens_completion
541
native_tokens_prompt
66
native_tokens_completion
361
native_tokens_completion_images
0
native_tokens_reasoning
42
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
0.00558
router
(null)
provider_responses
0
endpoint_id
"9ff5625c-403f-4d7f-b895-58ac7295062c"
id
"resp_0657280774493cee0169af4c944bc08193b17e48887275e821"
is_byok
false
latency
835
model_permaslug
"openai/gpt-5.4-20260305"
provider_name
"OpenAI"
status
200
user_agent
"langchainjs-openai/1.0.0 ((node/v24.14.0; linux; x64))"
http_referer
(null)
api_type
"completions"
id
"gen-1773096083-9Oxn8Aoj3QvULYK95Wo2"
upstream_id
"resp_0657280774493cee0169af4c944bc08193b17e48887275e821"
total_cost
0.00558
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenAI"
Evaluation details
Result
Evaluator
Details
Meta Data
68.10%
Matches word count
n/a
neededClean
false
words
214
98.43%
Dialogue to Total Word Ratio
Ratio: 93.55%, Deviation: 3.55%
neededClean
false
wordsTotal
217
wordsDialogue
203
83.2647%