NC Bench
Overview
About
Tests
Categories
Models
Vendors
Compare
Run: 2026-03-18-01KM0AEC6PMTN14F5M27RDE79G
Run Details
Model:
GPT-5.4 Mini (Reasoning, Low)
Test:
Dialogue tags
Scenario:
Write 200 words with 90% dialogue
Duration:
2.9s
Cost:
$0.001840
Provider:
OpenAI
Evaluation summary
Score
Evaluator
Details Link
14.30%
Matches word count
(details)
98.94%
Dialogue to Total Word Ratio
(details)
56.6209%
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, tell me again why we’re dressed as a clown and a magician,” Bill whispered. “Because, Bill, the party has a clown, a magician, and zero security.” “Tragic oversight.” “Delightful oversight,” Molly said. “We go in, smile at the parents, compliment the balloon giraffe, and leave with the presents.” “Under whose names?” “Yours, ideally. You have that trustworthy face.” “My face is a warning sign in a waistcoat.” “Exactly. Now hold this tray.” “This tray contains mini quiches.” “Correct. Nobody suspects a thief who respects finger food.” “What if someone asks what act we’re performing?” “You say, ‘The vanishing act.’ Then I distract them.” “With what?” “My excellent impression of a man who knows how to fold a napkin.” “That is criminal.” “Yes, and tonight we’re going to be terribly successful.” Bill peeked through the hedge. “There are children.” “Of course there are children. It’s a children’s party.” “Right. Just checking that the plan remains morally slippery.” “It does.” “And the presents?” “Molly grinned. “Stacked by the cake. Like someone invited destiny.” “Any chance we could steal only the expensive ones?” “Bill, we are professionals.” “We are wearing fake noses.” “Professional fake noses.” “And after?” “We split the haul, buy hats, and never speak of the clown car again.” “We don’t have a clown car.” “We will, if this goes well.”
created_at
"2026-03-18T11:14:10.276Z"
3/18/2026, 11:14:10 AM
model
"openai/gpt-5.4-mini-20260317"
app_id
182717
external_user
(null)
streamed
true
cancelled
false
latency
1927
moderation_latency
(null)
generation_time
2687
tokens_prompt
77
tokens_completion
445
native_tokens_prompt
66
native_tokens_completion
398
native_tokens_completion_images
0
native_tokens_reasoning
78
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
0.0018405
router
(null)
provider_responses
0
endpoint_id
"9ee065b2-3d1c-43bc-bdd7-28af3b148282"
id
"resp_0bda0c4e3f55d6190169ba890253ac819789a4b95113e83b54"
is_byok
false
latency
207
model_permaslug
"openai/gpt-5.4-mini-20260317"
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-1773832450-44RdkdZsZUathul2XnVd"
upstream_id
"resp_0bda0c4e3f55d6190169ba890253ac819789a4b95113e83b54"
total_cost
0.0018405
cache_discount
(null)
upstream_inference_cost
0
provider_name
"OpenAI"
Evaluation details
Result
Evaluator
Details
Meta Data
14.30%
Matches word count
n/a
neededClean
false
words
221
98.94%
Dialogue to Total Word Ratio
Ratio: 93.21%, Deviation: 3.21%
neededClean
false
wordsTotal
221
wordsDialogue
206
56.6209%