NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-09-01KKAC9G6JE8R3TDDW9PSVM8MN
Run Details
Model:
GPT-5.4 (Reasoning, Low)
Test:
Dialogue tags
Scenario:
Write 200 words with 10% dialogue
Duration:
16.3s
Cost:
$0.017625
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
100.00%
Matches word count
(details)
99.56%
Dialogue to Total Word Ratio
(details)
99.7809%
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
Molly crouched behind the hedge in a raincoat patterned like disappointed camouflage, studying the rented bouncy castle as if it were a bank vault with better balloons. Bill lay beside her, smelling faintly of jam and burglary, assembling their equipment: a cake server, two pillowcases, and a walkie-talkie stolen from a lifeguard. "Bill, no smashing, only smiling," Molly whispered. "Molly, if cornered, I can juggle and cry on command," Bill whispered back. The garden beyond fizzed with shrieks, paper hats, and parents trapped in folding chairs. At the center, a trestle table groaned under presents wrapped in dinosaurs, glitter, and enough tape to secure a submarine. Molly adjusted her clipboard, because every serious criminal looked more official with stationery. Bill practiced his harmless face, accidentally discovering a look usually associated with confused alpacas. Their plan was disgracefully simple: enter as entertainers, distract the sugar-addled mob with a magic trick involving a handkerchief and plausible shouting, then evacuate the gifts through the side gate where Bill had parked the van marked WINDOW CLEANING and, less helpfully, ICE CREAM. A clown inside twisted balloons. Molly sighed. Professional rivalry was everywhere. Somewhere, a piƱata swung in the breeze like a witness with secrets.
created_at
"2026-03-09T22:43:10.172Z"
3/9/2026, 10:43:10 PM
model
"openai/gpt-5.4-20260305"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
15644
moderation_latency
(null)
generation_time
16062
tokens_prompt
77
tokens_completion
452
native_tokens_prompt
66
native_tokens_completion
1164
native_tokens_completion_images
0
native_tokens_reasoning
887
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
0.017625
router
(null)
provider_responses
0
endpoint_id
"9ff5625c-403f-4d7f-b895-58ac7295062c"
id
"resp_093b17b202c51f700169af4cfe42e48196a124048f3213026e"
is_byok
false
latency
386
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-1773096190-SfjITXm16sKlNrTfbxqs"
upstream_id
"resp_093b17b202c51f700169af4cfe42e48196a124048f3213026e"
total_cost
0.017625
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
99.56%
Dialogue to Total Word Ratio
Ratio: 7.43%, Deviation: 2.57%
neededClean
false
wordsTotal
202
wordsDialogue
15
99.7809%