NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-09-01KKABYBC7VT0HGC77P1841YH0
Run Details
Model:
GPT-5.4 (Reasoning, Low)
Test:
Dialogue tags
Scenario:
Write 200 words with 50% dialogue
Duration:
24.6s
Cost:
$0.026460
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
100.00%
Matches word count
(details)
100.00%
Dialogue to Total Word Ratio
(details)
100.0000%
Overall
Model input
User/Human
You are a author of comedic heist books. Please write a 200 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
Molly spread a crayon-sketched floor plan across the dashboard of their idling van, tapping the kitchen, the cake table, and a suspiciously petting zoo. Bill adjusted his turtleneck and stared at the suburban house blazing with balloons. Inside, tiny guests ricocheted past windows like deranged confetti. A clown at the gate folded balloon swords with the grave concentration of a battlefield surgeon. The air smelled of frosting, juice boxes, and poor decisions. Molly checked her stopwatch; Bill checked whether remorse came in waves. It didn't, but indigestion did. Somewhere a child shrieked with delight, which both thieves misheard as opportunity. "Bill, this is not burglary; it's aggressive gift redistribution tonight." "Molly, we're robbing a clown guarded by six sugared-up toddlers." "Precisely, said Molly. No adult expects professionalism near a piƱata." "And the presents? whispered Bill. We take everything with ribbons?" "Everything, except the educational puzzles; we're thieves, not monsters, Bill." "What about the magician? He looks shifty, and employed professionally." "We neutralize him with applause, Molly said. Performers crave validation." "Then we vanish? asked Bill. Smoke bomb? Trampoline? Tasteful moonwalk?" "Vanish through the bouncy castle, Molly said. No footprints, dignity." "Fine, Bill sighed. If we're caught, I'm pretending to juggle."
created_at
"2026-03-09T22:37:04.786Z"
3/9/2026, 10:37:04 PM
model
"openai/gpt-5.4-20260305"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
22352
moderation_latency
(null)
generation_time
23883
tokens_prompt
77
tokens_completion
859
native_tokens_prompt
66
native_tokens_completion
1753
native_tokens_completion_images
0
native_tokens_reasoning
1454
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
0.02646
router
(null)
provider_responses
0
endpoint_id
"9ff5625c-403f-4d7f-b895-58ac7295062c"
id
"resp_0255ca7d5618f9ad0169af4b90df008195af4bc060b3dcbb8d"
is_byok
false
latency
181
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-1773095824-NQDu5CB12qLxTfrJAo6a"
upstream_id
"resp_0255ca7d5618f9ad0169af4b90df008195af4bc060b3dcbb8d"
total_cost
0.02646
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
100.00%
Dialogue to Total Word Ratio
Ratio: 50.00%, Deviation: 0.00%
neededClean
false
wordsTotal
202
wordsDialogue
101
100.0000%