NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-09-01KKAC0ZG66SBC6BDS8FGMG2TE
Run Details
Model:
GPT-5.4 (Reasoning, Low)
Test:
Dialogue tags
Scenario:
Write 200 words with 90% dialogue
Duration:
19.3s
Cost:
$0.017370
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
100.00%
Matches word count
(details)
66.12%
Dialogue to Total Word Ratio
(details)
83.0609%
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, this is our Mona Lisa,” Bill whispered. “Bouncy castle, distracted parents, and enough gift bags to retire on jelly beans.” “Bill, nobody retires on jelly beans.” “Metaphorically. We slip in as magicians, wave a wand, and vanish with the loot.” “You can’t do magic.” “I can do confidence, which is basically magic with pockets.” “The clown worries me.” “Every clown worries everyone. That’s his camouflage.” “What about the children?” “Short, sticky witnesses. We dazzle them with balloons.” “I hate balloons. They squeak like guilty shoes.” “Then face paint. No child suspects a woman with a butterfly on her cheek.” “I’m not being a butterfly.” “Dragon, then. Very intimidating.” “Bill, these presents are for a six-year-old.” “Exactly. Tiny electronics, premium chocolate, possibly cash in cards from lazy uncles.” “And the dog?” “We bribe him with sausage.” “What if there are two dogs?” “Then more sausage. Molly, this is logistics, not philosophy.” “If a child asks why we’re taking the presents?” “I’ll say, ‘Quality control.’” “And if a parent asks?” “You say, ‘We’re with entertainment.’” “What entertainment?” “Extremely disappointing magic.” Molly sighed. “Fine. But if I get glitter on my shoes, we’re leaving the television and absolutely not stealing the piñata.”
created_at
"2026-03-09T22:38:30.928Z"
3/9/2026, 10:38:30 PM
model
"openai/gpt-5.4-20260305"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
17227
moderation_latency
(null)
generation_time
19114
tokens_prompt
77
tokens_completion
390
native_tokens_prompt
66
native_tokens_completion
1147
native_tokens_completion_images
0
native_tokens_reasoning
846
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
0.01737
router
(null)
provider_responses
0
endpoint_id
"9ff5625c-403f-4d7f-b895-58ac7295062c"
id
"resp_0fbd6f3638d8d6c40169af4be6fe788196b9e4e84cef8d66d4"
is_byok
false
latency
232
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-1773095910-9VDDLXLCUmk33N6ss8OJ"
upstream_id
"resp_0fbd6f3638d8d6c40169af4be6fe788196b9e4e84cef8d66d4"
total_cost
0.01737
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenAI"
Evaluation details
Result
Evaluator
Details
Meta Data
100.00%
Matches word count
n/a
neededClean
false
words
200
66.12%
Dialogue to Total Word Ratio
Ratio: 98.02%, Deviation: 8.02%
neededClean
false
wordsTotal
202
wordsDialogue
198
83.0609%